CN115705474A - Document translation method, apparatus, storage medium, and program product - Google Patents

Abstract

Embodiments of the present disclosure provide a document translation method, device, storage medium, and program product. By analyzing the original document to be translated, at least one paragraph text to be translated and an initial syntax tree including the paragraph text are obtained, and the node mark where each paragraph text is located The identification of each paragraph copy; according to the style information of each paragraph copy, convert each paragraph copy into a paragraph copy with the first style tag; obtain the translation corresponding to each paragraph copy, and the translation is marked with the same position as the first style tag The second style label of the type; replace each paragraph copy in the initial syntax tree with its corresponding translation according to the identification of each paragraph copy, and restore the document according to the replaced syntax tree to obtain the translation document corresponding to the original document. By translating at the granularity of paragraph copy, the semantics of the translation are complete and accurate, and by replacing the paragraph copy with the translation in the syntax tree, it is ensured that the style of the translated document is consistent with the original document, improving the quality of document translation.

Description

Document translation method, device, storage medium and program product

technical field

The embodiments of the present disclosure relate to the field of computer technology, and in particular to a document translation method, device, storage medium and program product.

Background technique

Document translation is the text processing process of translating a document in one language into a document in another target language. With the continuous deepening of the globalization process and the rapid development of the Internet, the demand for document translation is increasingly emerging.

Some machine translation methods on the market extract the bottom copy of the document, and even convert the source format of the document into another format, such as Google’s document translation, and then perform machine translation. The entire translation process is limited by the minimum unit granularity of extraction and The accuracy of the translation engine may lead to some uncontrollable translation results, such as unsmooth translation sentences, or uncontrollable deviations in the format; if you want to avoid these problems, you can increase the minimum unit granularity judgment threshold, but it is limited by machine translation The grammatical order cannot be accurately restored, and the scene is limited to one sentence or a few short paragraphs of text. The final translation result may still be semantically inappropriate, or the semantic context may be lost.

Contents of the invention

Embodiments of the present disclosure provide a document translation method and a document translation method, device, storage medium, and program product, so as to improve the accuracy of document translation and preserve the document format.

In a first aspect, an embodiment of the present disclosure provides a document translation method, including:

Analyzing the original document to be translated, obtaining at least one paragraph text to be translated, and an initial syntax tree including the paragraph text, wherein the node where each paragraph text in the initial syntax tree is marked with the identification of each paragraph text;

converting each paragraph copy into a paragraph copy with a first style tag according to the style information of each paragraph copy;

Obtain the translation corresponding to each paragraph copy, wherein the position corresponding to the first style label in the translation is marked with the same type of second style label;

Each paragraph text in the initial syntax tree is replaced with its corresponding translation according to the identification of each paragraph text, and the document is restored according to the replaced syntax tree to obtain a translation document corresponding to the original document.

In a second aspect, an embodiment of the present disclosure provides a document translation device, including:

A parsing unit, configured to parse the original document to be translated, obtain at least one paragraph text to be translated, and an initial syntax tree including the paragraph text, wherein the node where each paragraph text in the initial syntax tree is marked with each paragraph text logo;

a conversion unit, configured to convert each paragraph copy into a paragraph copy with a first style tag according to the style information of each paragraph copy;

a translation unit, configured to obtain the translation corresponding to each paragraph copy, wherein the position corresponding to the first style label in the translation is marked with the same type of second style label;

The restoration unit is configured to replace each paragraph text in the initial syntax tree with its corresponding translation according to the identification of each paragraph text, and restore the document according to the replaced syntax tree to obtain the translation document corresponding to the original document.

In a third aspect, an embodiment of the present disclosure provides an electronic device, including: at least one processor and a memory;

the memory stores computer-executable instructions;

The at least one processor executes the computer-executed instructions stored in the memory, so that the at least one processor executes the document translation method described in the above first aspect and various possible designs of the first aspect.

In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium, where computer-executable instructions are stored in the computer-readable storage medium, and when the processor executes the computer-executable instructions, the above first aspect and the first Aspects of various possible designs of the described document translation method.

In the fifth aspect, the embodiments of the present disclosure provide a computer program product, including computer-executable instructions. When the processor executes the computer-executable instructions, the document translation described in the above first aspect and various possible designs of the first aspect is realized. method.

The document translation method, device, storage medium, and program product provided in this embodiment obtain at least one paragraph text to be translated and an initial syntax tree including the paragraph text by parsing the original document to be translated, wherein each of the initial syntax trees The node where the paragraph copy is located is marked with the identification of each paragraph copy; according to the style information of each paragraph copy, each paragraph copy is converted into a paragraph copy with the first style label; the translation corresponding to each paragraph copy is obtained, and the translation is the same as the first The corresponding position of the style tag is marked with a second style tag of the same type; replace each paragraph copy in the initial syntax tree with its corresponding translation according to the identification of each paragraph copy, and restore the document according to the replaced syntax tree to obtain The translation document corresponding to the original document. By translating at the granularity of paragraph copy, the semantics of the translation can be guaranteed to be complete and accurate, and by replacing the corresponding paragraph copy with the translation in the syntax tree, the restoration of the document style based on the syntax tree can be realized, ensuring that the style of the translated document is consistent with that of the original document. Improve document translation and translation quality.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description These are some embodiments of the present disclosure. Those skilled in the art can also obtain other drawings based on these drawings without any creative effort.

Figure 1a is an example diagram of the original document to be translated;

Figure 1b and Figure 1c are schematic diagrams of machine translation results in the prior art;

FIG. 2 is a schematic flowchart of a document translation method provided by an embodiment of the present disclosure;

FIG. 3 is a schematic flowchart of a document translation method provided by an embodiment of the present disclosure;

Fig. 4a is a schematic diagram of a paragraph text with a first style tag provided by an embodiment of the present disclosure;

Fig. 4b is a schematic diagram of a translation with a second style tag provided by an embodiment of the present disclosure;

Fig. 5a is a schematic diagram of paragraph copy and translation in the Office document translation process provided by an embodiment of the present disclosure;

Fig. 5b is a schematic diagram of a translation document corresponding to an Office document provided by an embodiment of the present disclosure;

FIG. 6 is a schematic diagram of paragraph copy and translation in the cloud document translation process provided by an embodiment of the present disclosure;

FIG. 7 is a schematic diagram of paragraph copy and translation in the process of translating a Markdown document provided by an embodiment of the present disclosure;

FIG. 8 is a structural block diagram of a document translation device provided by an embodiment of the present disclosure;

FIG. 9 is a schematic diagram of a hardware structure of an electronic device provided by an embodiment of the present disclosure.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present disclosure clearer, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments It is a part of the embodiments of the present disclosure, but not all of them. Based on the embodiments in the present disclosure, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present disclosure.

Document translation is the text processing process of translating a document in one language into a document in another target language. With the continuous deepening of the globalization process and the rapid development of the Internet, the demand for document translation is increasingly emerging.

Some machine translation methods on the market extract the bottom copy of the document, and even convert the source format of the document into another format, such as Google’s document translation, and then perform machine translation. The entire translation process is limited by the minimum unit granularity of extraction and The accuracy of the translation engine may lead to some uncontrollable translation results, such as unsmooth translation sentences, or uncontrollable deviations in the format; if you want to avoid these problems, you can increase the minimum unit granularity judgment threshold, but it is limited by machine translation The grammatical order cannot be accurately restored, and the scene is limited to one sentence or a few short paragraphs of text. The final translation result may still be semantically inappropriate, or the semantic context may be lost. Figure 1a shows the original document to be translated, and Figures 1b and 1c show the results of existing machine translation, which shows that the translation quality is poor.

In order to solve the above-mentioned technical problems, the embodiment of the present disclosure provides a document translation method, which analyzes the original document to be translated, and extracts the paragraph copy from the original document at the granularity of paragraphs, so as to ensure that the semantics of the translation is complete and accurate, and through the syntax tree in The translation replaces the corresponding paragraph copy, realizes the restoration of the document style based on the grammar tree, and ensures that the style of the translated document is consistent with the original document; and the embodiments of the present disclosure are applicable to Office documents, cloud documents, Markdown documents, etc., and are universal.

Specifically, analyze the original document to be translated, obtain at least one paragraph text to be translated, and an initial syntax tree including the paragraph text, wherein the node where each paragraph text in the initial syntax tree is marked with the identification of each paragraph text ; According to the style information of each paragraph copy, convert each paragraph copy into a paragraph copy with the first style tag; obtain the translation corresponding to each paragraph copy, wherein the position corresponding to the first style tag in the translation is marked with the same type of second style label; replace each paragraph copy in the initial syntax tree with its corresponding translation according to the identification of each paragraph copy, and restore the document according to the replaced syntax tree to obtain the translation document corresponding to the original document.

The document translation method of the present disclosure will be described in detail below with reference to specific embodiments and accompanying drawings.

Embodiment one

Referring to FIG. 2 , FIG. 2 is a schematic flowchart of a document translation method provided by an embodiment of the present disclosure. The method of this embodiment can be applied in a terminal device or a server, and the document translation method includes:

S201. Analyze the original document to be translated, and obtain at least one paragraph text to be translated and an initial syntax tree including the paragraph text, wherein the node where each paragraph text is located in the initial syntax tree is marked with an identifier of each paragraph text.

In this embodiment, the original document to be translated may be in document format such as Office document, cloud document, Markdown document, etc. Of course, it may also be in other document format, such as html, xliff, pdf, csv and other formats.

Considering that the accuracy of the existing machine translation translation process is limited by the minimum unit granularity extracted, and usually the minimum unit granularity extracted by machine translation is small, resulting in the loss of semantic context and inaccurate semantics, so this implementation In the example, paragraphs are used as the minimum unit granularity. The paragraph text obtained from the original document can be divided by dividing line breaks. A paragraph text may include one sentence or multiple sentences.

In this embodiment, the original document to be translated can be parsed, including splitting the original document, finding the paragraph text to be translated, and obtaining an initial syntax tree including the paragraph text, wherein the syntax tree can be an abstract syntax tree (AbstractSyntax Tree, AST), the abstract syntax tree is an abstract representation of the source code grammatical structure of the original document, and expresses the grammatical structure of the programming language in the form of a tree, and each node on the tree represents a structure in the source code. In this embodiment, when acquiring the initial syntax tree including the paragraph text, the identification of the paragraph text can be marked at the node where each paragraph text is located, so that the corresponding paragraph text can be located subsequently according to the identification of the paragraph text.

Optionally, as shown in Figure 3, when parsing the original document to be translated, obtaining at least one paragraph text to be translated and an initial syntax tree including the paragraph text, it may specifically include:

S2011. Determine the portion of the original document that includes the paragraph text, analyze the portion of the original document that includes the paragraph text, extract the paragraph text, and generate an initial syntax tree that includes the paragraph text;

S2012. Mark the identification of each paragraph copy in the node where each paragraph copy is located in the initial syntax tree;

S2013. Store the initial syntax tree.

In this embodiment, the original document usually consists of a series of data. For example, an Office document is essentially a Zip compression package, and its decompressed content is a folder composed of some xml files. Similarly, cloud documents and markdown documents also include a series of data, and in this embodiment only focus on the part that includes the paragraph copy to be translated, the original document can be disassembled to determine the part that includes the paragraph copy to be translated in the original document, and the part that includes the paragraph copy in the original document part of the analysis, extract the paragraph copy, and generate the initial syntax tree including the paragraph copy; and considering that the translation is not completed in an instant, a mapping identifier is required when replacing the paragraph copy with its corresponding translation, and these The logo is stored, so when parsing the original document to obtain the paragraph copy, mark the logo of each paragraph copy in the node where each paragraph copy is located in the initial syntax tree, and store the initial syntax tree marked with the logo of the paragraph copy as a temporary file, so that When the document is restored, the marked paragraph copy is extracted from the temporary file and replaced with the translated translation.

S202. According to the style information of each paragraph copy, convert each paragraph copy into a paragraph copy with a first style tag.

In this embodiment, since the paragraph text has a certain style, such as bold, italic, underline, strikethrough, background color, font color, etc., and may also include images, formulas, etc., the style information of the paragraph text can be identified, and based on The style information transforms the paragraph copy into a paragraph copy with the first style tag, as shown in Figure 4a.

Style tags can include double tags or single tags, where double tags include two tags, which are used to identify the start position and end position of the format, for example, for bold format, usually one or several characters are bolded, you can Insert a formatting label before the bold text, and insert a formatting label after the bold text, which is used to indicate that the text format between the two formatting labels is in bold format. A single tag is used to identify a format that exists independently, for example, it can be used to identify images, formulas, etc. In this embodiment, the style information of the paragraph copy can be obtained, and then the corresponding style tag can be inserted at the position corresponding to the style information in the paragraph copy. Insert the style tag corresponding to the bold format before the text content in the bold format, and insert the style tag corresponding to the bold format after the text content in the bold format.

In an optional embodiment, when converting each paragraph text into a paragraph text with a first style tag according to the style information of each paragraph text, it may specifically include:

According to the style information of each paragraph copy, determine the public style of the paragraph copy; according to the public style, identify the partial copy and/or target elements of the non-public style included in each paragraph copy, and correct the non-public style in the paragraph copy Add style tags corresponding to non-public styles for partial copy and/or target elements.

In this embodiment, the public style of the paragraph text can be determined first, where the public style can be the style that appears most in the paragraph text, for example, most of the text content in the paragraph text is italic, and only a small part of the text content is other style, italic is determined as the public style of the paragraph copy; or a certain style in the paragraph copy can be designated as the public style of the paragraph copy. Based on the public style, the content of the non-public style in the paragraph copy can be identified. The text content of the non-public style is the partial copy, and the image, formula, etc. are the target elements. Because they have spatial attributes, they can also be called images, Target elements such as formulas are partial placeholder elements, and then add style tags corresponding to their styles before and after the partial copy, and add style tags corresponding to their styles at the target element.

S203. Obtain the translation corresponding to each paragraph copy, wherein the position corresponding to the first style tag in the translation is marked with the same type of second style tag.

In this embodiment, obtaining the translation corresponding to each paragraph copy can be realized through manual translation or machine translation, wherein during manual translation, the paragraph copy is displayed on the terminal interface, and the translation corresponding to the paragraph copy input by the user is received; while machine translation is Input the paragraph copy into the machine translation tool, and output the translation of the paragraph copy.

Furthermore, the paragraph copy contains the first style tag, and it is necessary to insert the same type of second style tag at the corresponding position of the corresponding translation. For example, a part of the paragraph copy is in italic format, and the part of the copy is inserted before and after In order to identify the style tag in italic format, a style tag identifying italic format also needs to be inserted before and after the content corresponding to the partial copy in the translation, as shown in Figure 4b.

Optionally, when inserting a style tag in the translation, the user can input a tag insertion command. The tag insertion command can include the insertion position and the type of the tag to be inserted. The insertion position is the position in the translation corresponding to the first style tag of the paragraph copy. Insert The type of the tag is a second-style tag of the same type as the first-style tag, and then in response to the user's tag insertion instruction, a second-style tag of the same type can be inserted in the translation at a position corresponding to the first-style tag of the paragraph copy Label.

S204. Replace each paragraph text in the initial syntax tree with its corresponding translation according to the identification of each paragraph text, and restore the document according to the replaced syntax tree to obtain a translation document corresponding to the original document.

In this embodiment, after obtaining the translations corresponding to all the paragraphs to be translated, the document restoration process can be carried out. Based on the idea of stealing the beam, the paragraph copy is located in the initial grammar tree according to the paragraph copy identifier, and the paragraph copy is replaced by For the corresponding translation, the replaced syntax tree is obtained, and the document is restored according to the replaced syntax tree to obtain the translation document corresponding to the original document, while ensuring that the style of the translation document is consistent with the style of the original document.

Optionally, in S201, when parsing the part including the paragraph text and extracting the paragraph text, the paragraph text may be extracted from the part of the original document including the paragraph text to generate a first paragraph unit data list, wherein the first paragraph unit data The list is to record the paragraph copy that needs to be translated in the form of a list, and the paragraph copy is displayed in its style, as shown in Table 1:

Table 1

logo A B C paragraph copy Hello worker have you eaten? You! The other party sends you an invitation to cook a meal! Whether to accept Walk

Further, in S202, when converting each paragraph copy into a paragraph copy with a first style tag according to the style information of each paragraph copy, the paragraph can be determined according to the style information of each paragraph copy in the first paragraph unit data list. The public style of copywriting; according to the public style, identify the partial copywriting and/or target elements of the non-public style included in each paragraph copy in the first paragraph unit data list, and modify the partial copywriting of the non-public style in the paragraph copy Copywriting and/or target elements add style tags corresponding to non-public styles.

Further, when replacing each paragraph text in the initial grammar tree with its corresponding translation according to the identification of each paragraph text, the translation corresponding to each paragraph text can be converted into a second paragraph unit data list, the first The translation style in the two-paragraph unit data list corresponds to the corresponding paragraph copywriting style, as shown in Table 2:

Table 2

Further, each paragraph text in the initial syntax tree may be replaced with its corresponding translation according to the identification of each paragraph text and the second paragraph unit data list, and the identification of the paragraph text may be deleted.

Specifically, in finding the corresponding paragraph copy node, it is necessary to divide the paragraph copy exported from the paragraph editor into the partial copy and partial placeholder of the translation according to the style tags in the corresponding translation, and then traverse the first paragraph of the original paragraph copy interface in turn Partial copywriting or partial placeholders in the unit data list are replaced with translated data node information in turn, so that the paragraph copywriting extracted from the analysis is truly translated according to the translation requirements.

The document translation method provided in this embodiment obtains at least one paragraph text to be translated and an initial syntax tree including the paragraph text by parsing the original document to be translated, wherein the nodes where the texts of each paragraph in the initial syntax tree are marked with each paragraph Copywriting identification; according to the style information of each paragraph copywriting, each paragraph copywriting is converted into a paragraph copywriting with the first style tag; obtain the translation corresponding to each paragraph copywriting, wherein the position corresponding to the first style tag in the translation is marked with The second style tag of the same type; replace each paragraph copy in the initial syntax tree with its corresponding translation according to the identification of each paragraph copy, and restore the document according to the replaced syntax tree to obtain the translation document corresponding to the original document. By translating at the granularity of paragraph copy, the semantics of the translation can be guaranteed to be complete and accurate, and by replacing the corresponding paragraph copy with the translation in the syntax tree, the restoration of the document style based on the syntax tree can be realized, ensuring that the style of the translated document is consistent with that of the original document. Improve document translation and translation quality.

On the basis of the above-mentioned embodiments, the document translation method of the present disclosure will be respectively introduced below using three different document types: Office, cloud documents, and Markdown documents.

Embodiment two

On the basis of the foregoing embodiments, in this embodiment, an Office document is taken as an example to introduce a document translation method.

For Office documents, including Docx documents, Xlsx documents, etc., since Office documents are essentially a Zip archive, its decompressed content is a folder composed of some xml files, so the analysis and restoration of the document translation method in this embodiment only need Attention needs to pay attention to the xml file including the paragraph copy to be translated, extract the paragraph copy from it, translate and transform it, and finally generate the xml file of the translation corresponding to the original xml file, in which the relevant content of the paragraph copy has been replaced with the translation, Restore it to the original xml file location when parsing, and then compress it into a Zip archive.

According to the above process, the document translation method for Office documents in this embodiment includes the following processes: format dismantling, document parsing, translation, and document restoration. Each process is described in detail below.

2.1) Format dismantling

Format disassembly is the pre-process of document parsing, so as to determine the part of the original document that includes the paragraph copy to be translated.

For Docx documents, the folder directory after the Docx document is converted to Zip format and decompressed roughly includes: [Content_Types].xml, _rels/folder, docProps/folder, word/folder.

Among them, [Content_Types].xml: describe the ContentType of each part of the document (such as: docment.xml), so that the program knows how to parse the part when displaying the document.

The text files that need to be paid attention to in the analysis of Docx documents include the content body, headers, footers, endnotes, footnotes, text comments, etc., and xml files related to paragraph copy text. Therefore, these xml files including paragraph copy can be obtained after decompressing the Docx file.

For Xlsx documents, the folder directory after the Xlsx document is converted to Zip format and decompressed roughly includes: [Content_Types].xml, _rels/folder, docProps/folder, xl/folder

Among them, [Content_Types].xml: describe each part of the document, so that the program knows how to parse the part when displaying the document.

For the text inside the Xlsx document, first extract all the copywriting in the sheet table from the ShareStrings. The >index<v></c> subscript mapping method is injected into the text rendering process of sheet[n]. It should be noted that in some old Xlsx versions, there is no real text in SharesString.xml, and the text content directly exists in <c><v>content<v></c>. At this time, do not use subscripts map.

In this embodiment, the Docx and Xlsx can be disassembled and the paragraph texts can be extracted through the pre-process to generate the first paragraph unit data list, and the paragraph texts to be translated can be recorded in the form of a list. Specifically, it can be achieved through the following process:

When the Docx and Xlsx documents are passed in, they are passed to jszip in the form of Buffer (cache) for loading. After getting the parsed content of the file, you can see the parsed content (such as each xml file, etc.) with the path path zipObject(key,value ) form exists.

Further, traverse the parsed content and find the file that needs attention, that is, the part that includes the paragraph copy, as follows:

For Docx documents: document.xml, header1.xml, font1.xml and other xml files including paragraph copywriting to be translated;

For Xlsx documents: shareStrings.xml, sheet1.xml and other xml files that include paragraphs to be translated.

2.2) Document parsing

For Office documents, there is usually the concept of paragraph (paragraph), and the rendering principle is realized from the perspective of paragraph. For Xlsx and PPT documents, although there is no concept of paragraph, the underlying data is roughly the same. Therefore, when parsing documents, the granularity of extraction can be increased to the dimension of paragraphs to ensure complete and accurate semantics.

In this embodiment, it is necessary to pay attention to the style information of the paragraph copy and the partial copy corresponding to the style, so as to display it to the translator.

For different documents, the parsing method and the parsed data have certain differences. Therefore, different parsers need to be constructed when parsing, and the data that is similar in length but with different details can be parsed into general data. The input data of the converter is AST, and the output is general data. The general data is as follows:

The following describes how to obtain the required AST data.

Whether it is Docx or Xlsx, only the paragraph copy is concerned in the document parsing process, so it only needs to parse and convert the xml file including the paragraph copy to be translated into the AST data of the corresponding file.

In the process of parsing, the partial copy and partial placeholders included in each paragraph copy are traversed in order, so that the overall information of this part of the copy can be extracted, and the concept of tags can be used at the beginning and end of each traversal. A relative position record. And when parsing each paragraph copy, a paragraph copy identification information paragraghId is generated for it in the source data source (in the AST).

The processed AST data objects of the above-mentioned file types all have a common feature. They all use segmented paragraph copywriting paragraphs as copywriting units, which are composed of partial copywriting and partial placeholder elements that are distinguished by copywriting styles, as shown below .

Finally, the source data AST is stored as temporary data, which is called a temporary file here. The data extracted from AST is stored locally or remotely, so as to be translated based on these data.

Specifically, through the sax (simple API for XML) parser, the xml file including the paragraph copy can be converted into plain text and injected into the sax parser, and AST node information will be created when a tag element is encountered. AST is divided into element nodes and text node. A list of paragraph unit data from which paragraph copy can be extracted can be obtained, and these paragraph unit data can be traversed to be processed into identifiable paragraph copy (partial copy & partial placeholder tagging information). After obtaining these identifiable paragraph texts, the textRunList partial copy and local placeholders are converted into identifiable tag information.

Based on docx and xlsx itself splitting the local copy of each paragraph itself, extract the public styles on the style nodes that act on the partial copy, traverse the original split list, and mark the start and end positions of the copy except the public style Double tab. The pictures that are not copywriting will occupy the space, keep the original information, and put a single label on this position. Through the above process, the paragraph copy with the first style tag can be obtained, as shown in the left part of Fig. 5a.

2.3) Translation

After the document parsing process is executed, it can be seen that document parsing strongly depends on how the underlying data of the document disassembles the paragraph copy. That is to say, for the same style of paragraph copy, the partial copy label information obtained may be different. (For example, the Office series style considers fonts, characters, etc.), but no matter which format file will get a data object with a unified content format, the translation itself needs to completely translate the extracted paragraph copy.

In this embodiment, the goal of paragraph text translation is not only to translate the paragraph text itself into the language of the target language, but also to restore the original style (such as whether it is bold or italic) to the corresponding position, especially When there are various subtle subject-predicates.

The specific translation process can be realized through manual translation or machine translation. During manual translation, the paragraph copy is displayed on the terminal interface, and the translation corresponding to the paragraph copy entered by the user is received; while machine translation is to input the paragraph copy into the machine translation tool , output the translation of the paragraph copy.

Among them, manual translation can import data into a visual editor, and translators can flexibly translate each paragraph copy that has been parsed in the editor. During the translation process, they only need to pay attention to the parsed text. For the first style tag displayed in the paragraph copy data, just insert the same type of second style tag at the corresponding position of the translation, as shown in the right part of Figure 5a, the whole process does not need to pay attention to what the paragraph text originally contained Style, which can make the translator more focused on the translation copy.

Optionally, the translators can record the content of the translated paragraphs each time they translate, so that they can be reused in the subsequent translation process and improve the efficiency of translation. In addition, in the process of using the editor to translate, the editor also provides various translation engines, which is convenient for translators to quickly fill in the translation when translating some commonly used small phrases.

Optionally, there is a data record for each translation of the translator, so integrating a large amount of translation record data for machine training will better help improve the quality and efficiency of the translation copy.

2.4) Document restoration

When the translation process is completed, the translators have translated the paragraphs extracted from the document parsing process according to the requirements. At this time, the translation data can be converted into AST, and the text content in it has been replaced with the target language translation.

After the document parsing process is completed, not only the paragraph copy that needs to be translated is obtained, but also the paragraphId is recorded on the corresponding initial AST node during the extraction process, and the initial AST that records the paragraphId is saved as a temporary file, so it can be based on the paragraphId and the original document The text files in the format establish a one-to-one relationship, and a replaced AST is generated, in which the text content has been replaced with the target language, and finally the restoration process is performed according to the corresponding restoration method of document analysis.

Specifically, it can be achieved through the following schemes:

In finding the corresponding paragraph copy node, it is necessary to divide the paragraph copy exported from the paragraph editor into the partial copy and the local placeholder according to the style tags in the translation corresponding to the paragraph copy, and then traverse the first paragraph unit data list of the original paragraph copy interface in turn Partial copywriting or partial placeholders in , are replaced with translated data node information in turn, so that the paragraph copywriting extracted from the analysis is truly translated according to the translation requirements.

More specifically, after the paragraph copy is translated, the translation data with style tags is obtained, which needs to be converted into a second paragraph unit data list before docx and xlsx can recognize it. That is, it is necessary to restore the translation data with style tags to the content of the corresponding style according to the style tags. The translation data with style tags can be traversed. If a double tag is recognized, the text before the double tag is determined to be normal text, and it is set as a public style, and the text inside the double tag also has a double tag on the basis of the public style corresponding style. If a single label is identified, it is determined that there is a local placeholder element, and then the local placeholder is performed. If no style tag is recognized, the style is set to a public style.

Further, the restored second paragraph unit data list and the first result paragraph unit data list are matched according to the paragraphId, and the data in the textRunList is replaced (the js object is a reference, so it will directly act on the initial AST data), in addition When extracting the paragraph copy and generating the initial AST, the raw value records the paragraphId of the original node data, so the paragraphId can be deleted after restoring a paragraph unit data.

No matter whether it is Docx or Xlsx, during the parsing process, you are concerned about some xml files including paragraph copywriting. During the restoration process, you only need to recreate a new xml file from the translated AST, and then replace it with the extracted position. Unzip the entire folder and recompress it into a compressed package in zip format, and rename it to .docx or xlsx.

It should be noted that the requirements of the Docx format will be different from those of other documents. In Docx, fonts in different languages will be rendered in different ways. For example, Arabic is rendered from right to left. According to the target language of document translation, In the document restoration stage, the language font style of the target language can be injected into the partial copy.

Through the above process, a translated document as shown in Figure 5b can be obtained, the style of the translated document is consistent with that of the original document, and the translation quality is improved.

Embodiment Three

On the basis of the foregoing embodiments, in this embodiment, a cloud document is taken as an example to introduce a document translation method.

For cloud documents, for example, Lark cloud documents support real-time online collaborative editing by multiple people. The rich text data object of the original document can be obtained from the cloud document server, and can be processed by parsing Office documents. It should be noted that the cloud document does not directly insert the obtained data back intact, because some information elements are not directly created, and unsupported content needs to be discarded. Here, the unsupported information can be pre-set List the elements, and then create a forbidden list to filter the unsupported content, and then remove it when it is supported later.

According to the above process, the document translation method for cloud documents in this embodiment includes the following processes: requesting data, document parsing, translation, document restoration, and uploading data. Each process is described in detail below.

3.1) Request data

For cloud documents, when sending rich text data acquisition requests to the server, identity verification is performed first, authentication is performed through the authentication logic provided by the API (Application Programming Interface, application programming interface), and the token identification of the cloud documents is passed in, according to the token Identifies the rich text data obtained from the cloud document.

Specifically, first verify the application authorization credential code, and then verify the user identity user_access_token, and then carry the lark_doc_token for obtaining document information and the token in the header in the request, and finally obtain the rich text data of the document.

3.2) Document parsing

Usually, in the parsing and extraction of machine translation, you only need to pay attention to the extracted copy itself, not the style of the copy and the relationship between the text and the text. In this embodiment, you need to pay attention to the paragraph copy The style information of , as well as the partial copywriting corresponding to the style, are displayed to translators.

For different documents, the parsing method and the parsed data have certain differences. Therefore, different parsers need to be constructed when parsing, and the data that is similar in length but with different details can be parsed into general data. The input data of the converter is AST, and the output is general data. The general data is as follows:

For cloud documents, the rich text data object of the cloud document link is requested from the server. In a non-strict sense, the rich text data object can be simply understood as a pseudo AST,

In addition, by recording the rich text data object of the cloud document as the first paragraph unit data list, the recorded data is further converted into paragraph copy data that can be recognized by the paragraph editor, and the data is processed, and secondary processing is performed where the style exists (Same as Markdown, see Embodiment 4 for details), finally obtain the paragraph copy data with the first style tag, as shown in the left part of FIG. 6 .

3.3) Translation

The goal of paragraph copy translation is not only to translate the paragraph copy itself into the target language, but also to restore the original style (such as whether it is bold or italic) to the corresponding position, especially when there are various subtle when the subject predicate.

The specific translation process can be realized through manual translation or machine translation. During manual translation, the paragraph copy is displayed on the terminal interface, and the translation corresponding to the paragraph copy entered by the user is received; while machine translation is to input the paragraph copy into the machine translation tool , output the translation of the paragraph copy.

Among them, manual translation can import data into a visual editor, and translators can flexibly translate each paragraph copy that has been parsed in the editor. During the translation process, they only need to pay attention to the parsed text. For the first-style tag displayed in the paragraph copy data, just insert the same type of second-style tag at the corresponding position of the translation, as shown in the right part of Figure 6, the whole process does not need to pay attention to what the paragraph text originally contained Style, which can make the translator more focused on the translation copy. The translation process is the same as that in Embodiment 2, and will not be repeated here.

3.4) Document restoration

After obtaining the translation data including the second style tag after the translation of the paragraph copy, it needs to be converted into a second paragraph unit data list. That is, it is necessary to restore the translation data with style tags to the content of the corresponding style according to the style tags. The translation data with style tags can be traversed. If a double tag is recognized, the text before the double tag is determined to be normal text, and it is set as a public style, and the text inside the double tag also has a double tag on the basis of the public style corresponding style. If a single label is identified, it is determined that there is a local placeholder element, and then the local placeholder is performed. If no style tag is recognized, the style is set to a public style.

Finally, backfill the second paragraph unit data list data into the marked temporary rich text data object (that is, the above-mentioned pseudo AST) according to the paragraphId tag according to the first result paragraph unit data list, and remove the paragraphId tag to generate The new rich text data object. Specifically, it can be achieved in the following ways:

Send the new rich text data object to the cloud document creation interface provided by the server, so that the cloud document of the translation can be created based on the new rich text data object, and the cloud document link of the translation can be returned, and the cloud document of the translation can be obtained according to the cloud document link document.

Embodiment Four

On the basis of the foregoing embodiments, in this embodiment, a Markdown document is taken as an example to introduce a document translation method.

First of all, it needs to be explained that whether it is an Office document or a cloud document, the parsed paragraph copy data is traversed from the beginning to the end of the copy, that is to say, there is no nesting relationship between the partial copy styles, which is conducive to double Label creation. For Markdown documents, in the parsed paragraph copy data, partial copy styles may have nested relationships, so the document parsing stage requires additional logical processing to smooth out nested relationships, and in the document restoration stage to create nested relationships.

According to the above, the document translation method for Markdown documents in this embodiment includes the following processes: pre-process, document parsing, translation, and document restoration. Each process is described in detail below.

4.1) Pre-process

Markdown content is plain text following the respective Markdown specification. The plain text content of Markdown needs to be disassembled according to the token mark formulated under different specifications, and the paragraph copy to be translated is obtained.

For the following Markdown document samples:

#Hello hit worker

*have you eaten? **You** Nah! *The other party sends you a **cooked meal** invitation! whether to receive

Walk! [img](http://tosv.boe.byted.org/obj/tostest/dagongren.png)

The existence form of Markdown files is actually a single long string of strings. Through traversal, it is subjected to grammatical and lexical operations, such as marked and remark, from block parsing (find titles, paragraphs...) to inline parsing (specifically Parse the title, paragraph...), and finally generate the initial AST syntax tree, and identify the paragraghId of the paragraph copy.

After parsing through remark, the initial AST syntax tree can be obtained, and then the paragraph unit data list can be obtained. Since the initial AST syntax tree of Markdown itself has a nested relationship in the processing of paragraph copy styles, as in the above example of the Markdown document, "eat Are you? You!", the word "you" is bolded, and the bold style is nested in the italic style in the corresponding copy style. That is to say, the bold "you" node is nested in the children object of the italic node, and the two sides of the "you" node are still ordinary text copy nodes. When parsing, the nesting relationship of the children style is recorded. When restoring, you can return it as it is.

It should be noted that in the syntax corresponding to Markdown documents, only some styles support nesting, such as bold, italic, strikethrough, such as inline code, link pictures, etc., although they exist in the form of styles (that is, they will be recorded in the formatting data field), but nesting is not supported.

After parsing the AST data, there is a nesting relationship between partial copywriting and paragraph copywriting, which leads to the direct nesting of styles, that is, the partial copywriting cannot be accurately distinguished directly by the style of the copywriting. This layer of nesting relationship must be resolved before entering document parsing and outputting general data.

Therefore, after obtaining the AST syntax tree of Markdown, it is necessary to flatten the start and end positions of the style content in the paragraph copy. Optionally, the paragraph copy can be divided into a parent style part and a child style part according to the style nesting relationship, and different first style tags are marked for the parent style part and the child style part; wherein, the first style tag of the parent style part It is used to identify the style information of the parent style, and the first style tag in the child style part is used to identify the style information of the parent style and the child style.

4.2) Document parsing

Usually, in the parsing and extraction of machine translation, you only need to pay attention to the extracted copy itself, not the style of the copy and the relationship between the text and the text. In this embodiment, you need to pay attention to the paragraph copy The style information of , as well as the partial copywriting corresponding to the style, are displayed to translators.

For different documents, the parsing method and the parsed data have certain differences. Therefore, different parsers need to be constructed when parsing, and the data that is similar in length but with different details can be parsed into general data. The input data of the converter is AST, and the output is general data. The general data is as follows:

For Markdown documents, the paragraph copy to be translated extracted and recorded in the above process can be recorded in the first paragraph unit data list, and then the first paragraph unit data list can be processed, and the style of the partial copy in the corresponding paragraph copy can be recorded in the original position The upper style tag tag information. Specifically, it can be realized through the following process:

Disassemble a single partial copy, if all style tags do not exist, return in advance; otherwise continue the following process;

Obtain the public style of the paragraph copy; locate the non-public style part in the paragraph copy based on the public style;

Locate the non-public style part with a nested relationship in the non-public style part, split the part into a parent style part and a child style part according to the style nesting relationship, and mark the parent style part and the child style part with different first style tags ;The first style tag in the parent style part is used to identify the style information of the parent style, and the first style tag in the child style part is used to identify the style information of the parent style and the child style; !", the word "you" has a bold style nested in the italic style, the italic is the parent style, and the bold is the child style, and the "Have you eaten? You!" is divided into three parts, including "Have you eaten?" ", "You", "Nah!", the first style tags of the two parts "Have you eaten?" and "Nah!" are used to identify the italic style, and the first style tag of the "You" part Used to identify the italic and bold parts;

And for the part belonging to the style that does not support the nesting relationship, and/or the part belonging to the partial placeholder, the corresponding first style tag is inserted.

Through the above process, the paragraph copy data with the first style tag is finally obtained, as shown in the left part of FIG. 7 .

4.3) Translation

The goal of paragraph copy translation is not only to translate the paragraph copy itself into the target language, but also to restore the original style (such as whether it is bold or italic) to the corresponding position, especially when there are various subtle when the subject predicate.

The specific translation process can be realized through manual translation or machine translation. During manual translation, the paragraph copy is displayed on the terminal interface, and the translation corresponding to the paragraph copy entered by the user is received; while machine translation is to input the paragraph copy into the machine translation tool , output the translation of the paragraph copy.

Among them, manual translation can import data into a visual editor, and translators can flexibly translate each paragraph copy that has been parsed in the editor. During the translation process, they only need to pay attention to the parsed text. For the first-style tag displayed in the paragraph copy data, just insert the same type of second-style tag at the corresponding position of the translation, as shown in the right part of Figure 7, and the whole process does not need to pay attention to what the paragraph text originally contained Style, which can make the translator more focused on the translation copy. The translation process is the same as that in Embodiment 2, and will not be repeated here.

4.3) Document restoration

In the parsing process of Mardown files, it is different from parsing files in other formats, because there is a nesting relationship based on the copy styles in the paragraph copy, and the nesting relationship needs to be smoothed out. Also in the restoration process, just These nested information need to be restored. The method of restoration here can be based on the existence of double tags (and the same style information) before and after the copy, so as to obtain a new AST, and finally use this new AST to generate a translated Markdown file.

That is, according to the style nesting relationship, the identification of each paragraph copy, and the translation of each paragraph copy, replace each paragraph copy in the initial syntax tree with its corresponding translation, and restore the style nesting relationship to generate a replaced syntax tree ;Generate the Markdown document corresponding to the translation document according to the replaced syntax tree.

Specifically, after obtaining the translation data after the translation of the paragraph copy, the translation data with the second style tag information is recorded from left to right according to the character string, and the nesting of records that need to be restored during the document submission process style relationship. Optionally, document restoration can be implemented through the following process:

Traversing the translation data with the second style tag information, sequentially identifying each second style tag in the translation on the basis of the public style, locating the parts with nested relationship, and locating the parent style part and child style part, and then restoring the parent style The nesting relationship between styles and substyles;

For other second-style tags, revert to the corresponding style.

Further, restore the corresponding paragraph copy according to the temporary paragraphId information recorded when the Markdown document was submitted. Specifically, after restoring the nesting relationship, replace the paragraph copy in the initial syntax tree with the corresponding translation according to the paragraphId of each paragraph copy; and then generate a translated Markdown document according to the replaced syntax tree through the remark process.

After the above process, the content of the translated Markdown document is as follows:

#Hello commuter

*Have**you**eaten yet? *I would like to invite you**to dinner**! Accept the invitation?

Yes, with pleasure! [img](http://tosv.boe.byted.org/obj/tostest/dagongren.png)

It can be seen that compared with machine translation, the quality of document translation has been greatly improved after translation.

Embodiment five

Corresponding to the document translation method in the above embodiments, FIG. 8 is a structural block diagram of a document translation device provided in an embodiment of the present disclosure. For ease of description, only the parts related to the embodiments of the present disclosure are shown. Referring to FIG. 8 , the document translation device 800 includes: a parsing unit 801 , a conversion unit 802 , a translation unit 803 , and a restoration unit 804 .

Parsing unit 801, configured to parse the original document to be translated, obtain at least one paragraph text to be translated, and an initial syntax tree including the paragraph text, wherein the nodes where each paragraph text in the initial syntax tree is marked with each paragraph copywriting identification;

A conversion unit 802, configured to convert each paragraph copy into a paragraph copy with a first style tag according to the style information of each paragraph copy;

The translation unit 803 is configured to obtain the translation corresponding to each paragraph text, wherein the position corresponding to the first style label in the translation is marked with the same type of second style label;

The restoration unit 804 is configured to replace each paragraph text in the initial syntax tree with its corresponding translation according to the identification of each paragraph text, and perform document restoration according to the replaced syntax tree to obtain the translation document corresponding to the original document.

In a possible embodiment, when the parsing unit 801 parses the original document to be translated, and acquires at least one paragraph text to be translated and an initial syntax tree including the paragraph text, it is configured to:

determining the part of the original document that includes the paragraph text, parsing the part of the original document that includes the paragraph text, extracting the paragraph text, and generating an initial syntax tree including the paragraph text;

In the initial grammar tree, mark the identification of each paragraph copy at the node where each paragraph copy is located;

The initial syntax tree is stored.

In a possible embodiment, when the translation unit 803 obtains the translation corresponding to each paragraph text, it is used to:

Receive the translation corresponding to each paragraph copy input by the user or output by machine translation;

In response to a user's tag insertion instruction, a second style tag of the same type is inserted at a position in the translation corresponding to the first style tag of the paragraph copy.

In a possible embodiment, when parsing the part including the paragraph text and extracting the paragraph text, the parsing unit 801 is configured to:

extracting the paragraph text from the part of the original document that includes the paragraph text, and generating a first paragraph unit data list;

The conversion unit 802 is used to:

According to the style information of each paragraph copy in the first paragraph unit data list, determine the common style of the paragraph copy;

According to the public style, identify the partial copy and/or target elements of the non-public style included in each paragraph copy in the first paragraph unit data list, and perform the local copy and/or target element of the non-public style in the paragraph copy Elements add style tags corresponding to non-public styles.

In a possible embodiment, when replacing each paragraph text in the initial syntax tree with its corresponding translation according to the identification of each paragraph text, the restoration unit 804 is configured to:

Convert the translation corresponding to each paragraph copy into the second paragraph unit data list;

According to the identification of each paragraph text and the second paragraph unit data list, each paragraph text in the initial syntax tree is replaced with its corresponding translation, and the identification of the paragraph text is deleted.

In a possible embodiment, the parsing unit 801 determines the part of the original document that includes the paragraph text, parses the part of the original document that includes the paragraph text, extracts the paragraph text, and generates Initial syntax tree for paragraph copy, including:

If the original document is an Office document, after decompressing the compressed package of the Office document, an xml file including paragraph text is obtained;

The xml file including the paragraph text is parsed by a sax parser, the paragraph text is extracted, and the initial syntax tree is generated.

In a possible embodiment, the restoring unit 804 is configured to:

Create a new xml file based on the replaced syntax tree;

Replace the xml file containing the paragraph text in the decompressed file of the Office document zip package with the new xml file;

Recompress the decompressed files of the Office document compression package to obtain the Office document compression package corresponding to the translated document.

In a possible embodiment, the parsing unit 801 determines the portion of the original document that includes the paragraph text, parses the portion of the original document that includes the paragraph text, extracts the paragraph text, and generates an When copywriting the initial syntax tree, use to:

If the original document is a cloud document, send a rich text data acquisition request to the server, and receive the rich text data object of the cloud document sent by the server after authentication;

The paragraph copy is obtained according to the rich text object, and the rich text data object is determined as the initial syntax tree.

In a possible embodiment, the restoration unit 804 replaces each paragraph text in the initial syntax tree with its corresponding translation according to the identification of each paragraph text, and restores the document according to the replaced syntax tree, When obtaining the translation document corresponding to the original document, it is used to:

Generate a rich text data object of the translation according to the identification of each paragraph copy, the translation of each paragraph copy and the initial syntax tree;

Sending the rich text data object of the translation to the server, so that the server creates a cloud document of the translation according to the rich text data object of the translation, and returns the cloud document link of the translation;

Obtain the cloud document of the translation according to the cloud document link.

In a possible embodiment, the parsing unit 801 determines the part of the original document that includes the paragraph text, parses the part of the original document that includes the paragraph text, extracts the paragraph text, and generates When initial syntax tree for paragraph copy, use for:

If the original document is a Markdown document, block parsing and inline parsing are performed on the Markdown document to obtain a paragraph copy and generate an initial syntax tree.

In a possible embodiment, when the conversion unit 802 converts each paragraph text into a paragraph text with a first style tag according to the style information of each paragraph text, it is used to:

Obtaining the style nesting relationship in the paragraph copy according to the initial syntax tree;

dividing the paragraph copy into a parent style part and a child style part according to the style nesting relationship, and marking the parent style part and the child style part with different first style tags;

Wherein, the first style tag in the parent style part is used to identify the style information of the parent style, and the first style tag in the child style part is used to identify the style information of the parent style and the child style.

In a possible embodiment, the restoration unit 804 replaces each paragraph text in the initial syntax tree with its corresponding translation according to the identification of each paragraph text, and restores the document according to the replaced syntax tree, When obtaining the translation document corresponding to the original document, it is used to:

According to the style nesting relationship, the identification of each paragraph copy, and the translation of each paragraph copy, replace each paragraph copy in the initial syntax tree with its corresponding translation, restore the style nesting relationship, and generate a replaced grammar Tree;

Generate the Markdown document corresponding to the translation document according to the replaced syntax tree.

The device provided in this embodiment can be used to implement the technical solution of the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not repeat them here.

Embodiment six

Referring to FIG. 9 , it shows a schematic structural diagram of an electronic device 900 suitable for implementing the embodiments of the present disclosure. The electronic device 900 may be a terminal device or a server. Wherein, the terminal equipment may include but not limited to mobile phones, notebook computers, digital broadcast receivers, personal digital assistants (Personal Digital Assistant, PDA for short), tablet computers (Portable Android Device, PAD for short), portable multimedia players (Portable MediaPlayer (PMP for short), mobile terminals such as vehicle-mounted terminals (such as vehicle-mounted navigation terminals), and fixed terminals such as digital TVs and desktop computers. The electronic device shown in FIG. 9 is only an example, and should not limit the functions and application scope of the embodiments of the present disclosure.

As shown in FIG. 9, an electronic device 900 may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 901, which may be stored in a program in a read-only memory (Read Only Memory, ROM for short) 902 or from a storage device. 908 programs loaded into the Random Access Memory (RAM for short) 903 to execute various appropriate actions and processes. In the RAM 903, various programs and data necessary for the operation of the electronic device 900 are also stored. The processing device 901 , ROM 902 , and RAM 903 are connected to each other through a bus 904 . An input/output (I/O) interface 905 is also connected to the bus 904 .

Generally, the following devices can be connected to the I/O interface 905: an input device 906 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; including, for example, a Liquid Crystal Display (LCD for short) , an output device 907 such as a speaker, a vibrator, etc.; a storage device 908 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 909. The communication means 909 may allow the electronic device 900 to perform wireless or wired communication with other devices to exchange data. While FIG. 9 shows electronic device 900 having various means, it is to be understood that implementing or having all of the means shown is not a requirement. More or fewer means may alternatively be implemented or provided.

In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, where the computer program includes program codes for executing the methods shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from a network via communication means 909 , or from storage means 908 , or from ROM 902 . When the computer program is executed by the processing device 901, the above-mentioned functions defined in the methods of the embodiments of the present disclosure are performed.

It should be noted that the above-mentioned computer-readable medium in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. A computer readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, electrical connections with one or more wires, portable computer diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable Programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present disclosure, however, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave carrying computer-readable program code therein. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can transmit, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device . Program code embodied on a computer readable medium may be transmitted by any appropriate medium, including but not limited to wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

The above-mentioned computer-readable medium may be included in the above-mentioned electronic device, or may exist independently without being incorporated into the electronic device.

The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by the electronic device, the electronic device is made to execute the methods shown in the above-mentioned embodiments.

Computer program code for carrying out the operations of the present disclosure can be written in one or more programming languages, or combinations thereof, including object-oriented programming languages—such as Java, Smalltalk, C++, and conventional Procedural Programming Language - such as "C" or a similar programming language. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In cases involving a remote computer, the remote computer can be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or it can be connected to an external computer (e.g. using an Internet Service Provider to connect via the Internet).

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or portion of code that contains one or more logical functions for implementing specified executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved. It should also be noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by a dedicated hardware-based system that performs the specified functions or operations , or may be implemented by a combination of dedicated hardware and computer instructions.

The units involved in the embodiments described in the present disclosure may be implemented by software or by hardware. Wherein, the name of the unit does not constitute a limitation of the unit itself under certain circumstances, for example, the first obtaining unit may also be described as "a unit for obtaining at least two Internet Protocol addresses".

The functions described herein above may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), System on Chips (SOCs), Complex Programmable Logical device (CPLD) and so on.

In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include one or more wire-based electrical connections, portable computer discs, hard drives, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, compact disk read only memory (CD-ROM), optical storage, magnetic storage, or any suitable combination of the foregoing.

In a first aspect, according to one or more embodiments of the present disclosure, a document translation method is provided, including:

Analyzing the original document to be translated, obtaining at least one paragraph text to be translated, and an initial syntax tree including the paragraph text, wherein the node where each paragraph text in the initial syntax tree is marked with the identification of each paragraph text;

converting each paragraph copy into a paragraph copy with a first style tag according to the style information of each paragraph copy;

Obtain the translation corresponding to each paragraph copy, wherein the position corresponding to the first style label in the translation is marked with the same type of second style label;

Each paragraph text in the initial syntax tree is replaced with its corresponding translation according to the identification of each paragraph text, and the document is restored according to the replaced syntax tree to obtain a translation document corresponding to the original document.

According to one or more embodiments of the present disclosure, the parsing the original document to be translated, obtaining at least one paragraph text to be translated and an initial syntax tree including the paragraph text includes:

determining the part of the original document that includes the paragraph text, parsing the part of the original document that includes the paragraph text, extracting the paragraph text, and generating an initial syntax tree including the paragraph text;

In the initial grammar tree, mark the identification of each paragraph copy at the node where each paragraph copy is located;

The initial syntax tree is stored.

According to one or more embodiments of the present disclosure, the obtaining the translation corresponding to each paragraph text includes:

Receive the translation corresponding to each paragraph copy input by the user or output by machine translation;

In response to a user's tag insertion instruction, a second style tag of the same type is inserted at a position in the translation corresponding to the first style tag of the paragraph copy.

According to one or more embodiments of the present disclosure, the parsing the part including the paragraph text, and extracting the paragraph text includes:

extracting the paragraph text from the part of the original document that includes the paragraph text, and generating a first paragraph unit data list;

The step of converting each paragraph copy into a paragraph copy with a first style tag according to the style information of each paragraph copy includes:

According to the style information of each paragraph copy in the first paragraph unit data list, determine the common style of the paragraph copy;

According to the public style, identify the partial copy and/or target elements of the non-public style included in each paragraph copy in the first paragraph unit data list, and perform the local copy and/or target element of the non-public style in the paragraph copy Elements add style tags corresponding to non-public styles.

According to one or more embodiments of the present disclosure, the replacing each paragraph text in the initial syntax tree with its corresponding translation according to the identification of each paragraph text includes:

Convert the translation corresponding to each paragraph copy into the second paragraph unit data list;

According to the identification of each paragraph text and the second paragraph unit data list, each paragraph text in the initial syntax tree is replaced with its corresponding translation, and the identification of the paragraph text is deleted.

According to one or more embodiments of the present disclosure, determining the part of the original document that includes the paragraph text, parsing the part of the original document that includes the paragraph text, extracting the paragraph text, and generating The initial syntax tree of , including:

If the original document is an Office document, after decompressing the compressed package of the Office document, an xml file including paragraph text is obtained;

The xml file including the paragraph text is parsed by a sax parser, the paragraph text is extracted, and the initial syntax tree is generated.

According to one or more embodiments of the present disclosure, the document restoration according to the replaced syntax tree to obtain the translation document corresponding to the original document includes:

Create a new xml file based on the replaced syntax tree;

Replace the xml file containing the paragraph text in the decompressed file of the Office document zip package with the new xml file;

Recompress the decompressed files of the Office document compression package to obtain the Office document compression package corresponding to the translated document.

According to one or more embodiments of the present disclosure, determining the part of the original document that includes the paragraph text, parsing the part of the original document that includes the paragraph text, extracting the paragraph text, and generating The initial syntax tree of , including:

If the original document is a cloud document, send a rich text data acquisition request to the server, and receive the rich text data object of the cloud document sent by the server after authentication;

The paragraph copy is obtained according to the rich text object, and the rich text data object is determined as the initial syntax tree.

According to one or more embodiments of the present disclosure, the text of each paragraph in the initial syntax tree is replaced with its corresponding translation according to the identification of each paragraph text, and the document is restored according to the replaced syntax tree to obtain the original The translation document corresponding to the document, including:

Generate a rich text data object of the translation according to the identification of each paragraph copy, the translation of each paragraph copy and the initial syntax tree;

Sending the rich text data object of the translation to the server, so that the server creates a cloud document of the translation according to the rich text data object of the translation, and returns the cloud document link of the translation;

Obtain the cloud document of the translation according to the cloud document link.

According to one or more embodiments of the present disclosure, determining the part of the original document that includes the paragraph text, parsing the part of the original document that includes the paragraph text, extracting the paragraph text, and generating The initial syntax tree of , including:

If the original document is a Markdown document, block parsing and inline parsing are performed on the Markdown document to obtain a paragraph copy and generate an initial syntax tree.

According to one or more embodiments of the present disclosure, converting each paragraph text into a paragraph text with a first style tag according to the style information of each paragraph text includes:

Obtaining the style nesting relationship in the paragraph copy according to the initial syntax tree;

dividing the paragraph copy into a parent style part and a child style part according to the style nesting relationship, and marking the parent style part and the child style part with different first style tags;

Wherein, the first style tag in the parent style part is used to identify the style information of the parent style, and the first style tag in the child style part is used to identify the style information of the parent style and the child style.

According to one or more embodiments of the present disclosure, the text of each paragraph in the initial syntax tree is replaced with its corresponding translation according to the identification of each paragraph text, and the document is restored according to the replaced syntax tree to obtain the original The translation document corresponding to the document, including:

According to the style nesting relationship, the identification of each paragraph copy, and the translation of each paragraph copy, replace each paragraph copy in the initial syntax tree with its corresponding translation, restore the style nesting relationship, and generate a replaced grammar Tree;

Generate the Markdown document corresponding to the translation document according to the replaced syntax tree.

In a second aspect, according to one or more embodiments of the present disclosure, a document translation device is provided, including:

A parsing unit, configured to parse the original document to be translated, obtain at least one paragraph text to be translated, and an initial syntax tree including the paragraph text, wherein the node where each paragraph text in the initial syntax tree is marked with each paragraph text logo;

a conversion unit, configured to convert each paragraph copy into a paragraph copy with a first style tag according to the style information of each paragraph copy;

a translation unit, configured to obtain the translation corresponding to each paragraph copy, wherein the position corresponding to the first style label in the translation is marked with the same type of second style label;

The restoration unit is configured to replace each paragraph text in the initial syntax tree with its corresponding translation according to the identification of each paragraph text, and restore the document according to the replaced syntax tree to obtain the translation document corresponding to the original document.

According to one or more embodiments of the present disclosure, when the parsing unit parses the original document to be translated, and acquires at least one paragraph text to be translated and an initial syntax tree including the paragraph text, it is configured to:

determining the part of the original document that includes the paragraph text, parsing the part of the original document that includes the paragraph text, extracting the paragraph text, and generating an initial syntax tree including the paragraph text;

In the initial grammar tree, mark the identification of each paragraph copy at the node where each paragraph copy is located;

The initial syntax tree is stored.

According to one or more embodiments of the present disclosure, when acquiring the translation corresponding to each paragraph text, the translation unit is configured to:

Receive the translation corresponding to each paragraph copy input by the user or output by machine translation;

In response to a user's tag insertion instruction, a second style tag of the same type is inserted at a position in the translation corresponding to the first style tag of the paragraph copy.

According to one or more embodiments of the present disclosure, when the parsing unit parses the part including the paragraph text and extracts the paragraph text, it is used to:

extracting the paragraph text from the part of the original document that includes the paragraph text, and generating a first paragraph unit data list;

When the conversion unit converts each paragraph copy into a paragraph copy with a first style tag according to the style information of each paragraph copy, it is used for:

According to the style information of each paragraph copy in the first paragraph unit data list, determine the common style of the paragraph copy;

According to the public style, identify the partial copy and/or target elements of the non-public style included in each paragraph copy in the first paragraph unit data list, and perform the local copy and/or target element of the non-public style in the paragraph copy Elements add style tags corresponding to non-public styles.

According to one or more embodiments of the present disclosure, when the restoring unit replaces each paragraph text in the initial syntax tree with its corresponding translation according to the identification of each paragraph text, it is configured to:

Convert the translation corresponding to each paragraph copy into the second paragraph unit data list;

According to the identification of each paragraph text and the second paragraph unit data list, each paragraph text in the initial syntax tree is replaced with its corresponding translation, and the identification of the paragraph text is deleted.

According to one or more embodiments of the present disclosure, the parsing unit determines the portion of the original document that includes the paragraph text, parses the portion of the original document that includes the paragraph text, extracts the paragraph text, and generates The initial syntax tree of the above paragraph copy, including:

If the original document is an Office document, after decompressing the compressed package of the Office document, an xml file including paragraph text is obtained;

The xml file including the paragraph text is parsed by a sax parser, the paragraph text is extracted, and the initial syntax tree is generated.

According to one or more embodiments of the present disclosure, when the restoration unit restores the document according to the replaced syntax tree to obtain the translation document corresponding to the original document, it is configured to:

Create a new xml file based on the replaced syntax tree;

Replace the xml file containing the paragraph text in the decompressed file of the Office document zip package with the new xml file;

Recompress the decompressed files of the Office document compression package to obtain the Office document compression package corresponding to the translated document.

According to one or more embodiments of the present disclosure, the parsing unit determines the part of the original document that includes the paragraph text, parses the part of the original document that includes the paragraph text, extracts the paragraph text, and generates an When initial syntax tree for paragraph copy, use for:

If the original document is a cloud document, send a rich text data acquisition request to the server, and receive the rich text data object of the cloud document sent by the server after authentication;

The paragraph copy is obtained according to the rich text object, and the rich text data object is determined as the initial syntax tree.

According to one or more embodiments of the present disclosure, the restoration unit replaces each paragraph text in the initial syntax tree with its corresponding translation according to the identification of each paragraph text, and performs document restoration according to the replaced syntax tree , when obtaining the translation document corresponding to the original document, it is used for:

Generate a rich text data object of the translation according to the identification of each paragraph copy, the translation of each paragraph copy and the initial syntax tree;

Sending the rich text data object of the translation to the server, so that the server creates a cloud document of the translation according to the rich text data object of the translation, and returns the cloud document link of the translation;

Obtain the cloud document of the translation according to the cloud document link.

According to one or more embodiments of the present disclosure, the parsing unit determines the portion of the original document that includes the paragraph text, parses the portion of the original document that includes the paragraph text, extracts the paragraph text, and generates When describing the initial syntax tree of a paragraph copy, use it to:

If the original document is a Markdown document, block parsing and inline parsing are performed on the Markdown document to obtain a paragraph copy and generate an initial syntax tree.

According to one or more embodiments of the present disclosure, when converting each paragraph text into a paragraph text with a first style tag according to the style information of each paragraph text, the conversion unit is configured to:

Obtaining the style nesting relationship in the paragraph copy according to the initial syntax tree;

dividing the paragraph copy into a parent style part and a child style part according to the style nesting relationship, and marking the parent style part and the child style part with different first style tags;

Wherein, the first style tag in the parent style part is used to identify the style information of the parent style, and the first style tag in the child style part is used to identify the style information of the parent style and the child style.

According to one or more embodiments of the present disclosure, the restoration unit replaces each paragraph text in the initial syntax tree with its corresponding translation according to the identification of each paragraph text, and performs document restoration according to the replaced syntax tree , when obtaining the translation document corresponding to the original document, it is used for:

According to the style nesting relationship, the identification of each paragraph copy, and the translation of each paragraph copy, replace each paragraph copy in the initial syntax tree with its corresponding translation, restore the style nesting relationship, and generate a replaced grammar Tree;

Generate the Markdown document corresponding to the translation document according to the replaced syntax tree.

In a third aspect, according to one or more embodiments of the present disclosure, an electronic device is provided, including: at least one processor and a memory;

the memory stores computer-executable instructions;

The at least one processor executes the computer-executed instructions stored in the memory, so that the at least one processor executes the document translation method described in the above first aspect and various possible designs of the first aspect.

In a fourth aspect, according to one or more embodiments of the present disclosure, a computer-readable storage medium is provided, the computer-readable storage medium stores computer-executable instructions, and when a processor executes the computer-executable instructions, Realize the document translation method described in the above first aspect and various possible designs of the first aspect.

In the fifth aspect, according to one or more embodiments of the present disclosure, there is provided a computer program product, including computer-executable instructions, when the processor executes the computer-executable instructions, the above first aspect and various aspects of the first aspect are realized. Possible design of the described document translation method.

The above description is only a preferred embodiment of the present disclosure and an illustration of the applied technical principle. Those skilled in the art should understand that the disclosure scope involved in this disclosure is not limited to the technical solution formed by the specific combination of the above-mentioned technical features, but also covers the technical solutions formed by the above-mentioned technical features or Other technical solutions formed by any combination of equivalent features. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in this disclosure.

In addition, while operations are depicted in a particular order, this should not be understood as requiring that the operations be performed in the particular order shown or performed in sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Likewise, while the above discussion contains several specific implementation details, these should not be construed as limitations on the scope of the disclosure. Certain features that are described in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable subcombination.

Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are merely example forms of implementing the claims.

Claims (16)

1. A method of document translation, comprising:

analyzing an original document to be translated to obtain at least one paragraph pattern to be translated and an initial syntax tree comprising the paragraph pattern, wherein the node of each paragraph pattern in the initial syntax tree is marked with the identifier of each paragraph pattern;

converting each paragraph pattern into a paragraph pattern with a first pattern label according to the pattern information of each paragraph pattern;

acquiring a translation corresponding to each paragraph of the file, wherein a second style label with the same type is marked at a position in the translation corresponding to the first style label;

and replacing each paragraph and text in the initial grammar tree with a corresponding translation thereof according to the identification of each paragraph and text, and performing document restoration according to the replaced grammar tree to obtain a translation document corresponding to the original document.

2. The method of claim 1, wherein parsing the original document to be translated to obtain at least one paragraph pattern to be translated and an initial syntax tree including the paragraph pattern comprises:

determining a part of the original document including the paragraph pattern, analyzing the part of the original document including the paragraph pattern, extracting the paragraph pattern, and generating an initial syntax tree including the paragraph pattern;

marking the mark of each paragraph pattern on the node of each paragraph pattern in the initial syntax tree;

storing the initial syntax tree.

3. The method according to claim 2, wherein the obtaining of the translation corresponding to each paragraph of the document comprises:

receiving a translation corresponding to each paragraph of the document input by a user or output by machine translation;

and inserting a second style label of the same type at a position corresponding to the first style label of the paragraph file in the translation in response to the label inserting instruction of the user.

4. The method of claim 2, wherein parsing the portion including the paragraph pattern to extract the paragraph pattern comprises:

extracting paragraph patterns from the part of the original document including the paragraph patterns to generate a first paragraph unit data list;

converting each paragraph pattern into a paragraph pattern with a first pattern label according to the pattern information of each paragraph pattern, including:

determining the common style of the paragraph patterns according to the style information of each paragraph pattern in the first paragraph unit data list;

according to the common style, identifying a local style and/or a target element of a non-common style in each paragraph style in the first paragraph unit data list, and adding a style label corresponding to the non-common style to the local style and/or the target element of the non-common style in the paragraph style.

5. The method of claim 4, wherein said replacing each paragraph in the initial syntax tree with its corresponding translation based on the identification of each paragraph, comprises:

converting the translation corresponding to each paragraph case into a second paragraph unit data list;

and replacing each paragraph pattern in the initial syntax tree with a corresponding translation thereof according to the identifier of each paragraph pattern and the second paragraph unit data list, and deleting the identifier of the paragraph pattern.

6. The method of any one of claims 2-5, wherein determining the portion of the document that includes the paragraph pattern, parsing the portion of the document that includes the paragraph pattern, extracting the paragraph pattern, and generating the initial syntax tree that includes the paragraph pattern comprises:

if the original document is an Office document, decompressing an Office document compressed packet and then acquiring an xml file comprising paragraph patterns;

and analyzing the xml file comprising the paragraph and the text by using a sax analyzer, extracting the paragraph and the text and generating the initial syntax tree.

7. The method according to claim 6, wherein the performing document restoration according to the replaced syntax tree to obtain a translation document corresponding to the original document comprises:

creating a new xml file according to the replaced syntax tree;

replacing the xml file comprising the paragraph pattern in the Office document compressed packet decompression file with a new xml file;

and re-compressing the Office document compression packet decompression file to obtain an Office document compression packet corresponding to the translated document.

8. The method of any one of claims 2-5, wherein determining the portion of the document that includes the paragraph pattern, parsing the portion of the document that includes the paragraph pattern, extracting the paragraph pattern, and generating the initial syntax tree that includes the paragraph pattern comprises:

if the original document is a cloud document, sending a rich text data acquisition request to a server, and receiving a rich text data object of the cloud document sent by the server after authentication;

and acquiring the paragraph pattern according to the rich text object, and determining the rich text data object as the initial syntax tree.

9. The method of claim 8, wherein the replacing each paragraph in the initial syntax tree with its corresponding translation according to the identifier of each paragraph, and performing document restoration according to the replaced syntax tree to obtain a translation document corresponding to the original document, comprises:

generating a rich text data object of the translation according to the identification of each paragraph case, the translation of each paragraph case and the initial syntax tree;

sending the rich text data object of the translation to the server, so that the server creates a cloud document of the translation according to the rich text data object of the translation, and returns a cloud document link of the translation;

and acquiring the cloud document of the translation according to the cloud document link.

10. The method of any one of claims 2-5, wherein determining the portion of the document that includes the paragraph pattern, parsing the portion of the document that includes the paragraph pattern, extracting the paragraph pattern, and generating the initial syntax tree that includes the paragraph pattern comprises:

and if the original document is a Markdown document, performing block analysis and inline analysis on the Markdown document to obtain a paragraph pattern, and generating an initial syntax tree.

11. The method of any one of claim 10, wherein converting each paragraph document into a paragraph document having a first style label based on the style information of each paragraph document comprises:

acquiring a style nesting relation in the paragraph pattern according to the initial syntax tree;

segmenting the paragraph pattern into a parent pattern part and a child pattern part according to a pattern nesting relation, and marking different first pattern labels on the parent pattern part and the child pattern part;

the first style tag of the parent style part is used for identifying style information of the parent style, and the first style tag of the child style part is used for identifying style information of the parent style and the child style.

12. The method of claim 11, wherein the replacing each paragraph in the initial syntax tree with its corresponding translation according to the identifier of each paragraph, and performing document restoration according to the replaced syntax tree to obtain a translation document corresponding to the original document, comprises:

replacing each paragraph and text in the initial syntax tree with a corresponding translation thereof according to the style nesting relationship, the identification of each paragraph and text and the translation of each paragraph and text, and restoring the style nesting relationship to generate a replaced syntax tree;

and generating a Markdown document corresponding to the translation document according to the replaced syntax tree.

13. A document translation apparatus characterized by comprising:

the system comprises an analysis unit, a translation unit and a translation unit, wherein the analysis unit is used for analyzing an original document to be translated to obtain at least one paragraph pattern to be translated and an initial syntax tree comprising the paragraph pattern, and the node of each paragraph pattern in the initial syntax tree is marked with the identifier of each paragraph pattern;

the conversion unit is used for converting each paragraph pattern into the paragraph pattern with a first pattern label according to the pattern information of each paragraph pattern;

the translation unit is used for acquiring a translation corresponding to each paragraph and document, wherein the position of the translation corresponding to the first style label is marked with a second style label of the same type;

and the restoring unit is used for replacing each paragraph pattern in the initial grammar tree with the corresponding translation thereof according to the identification of each paragraph pattern, and restoring the document according to the replaced grammar tree to obtain the translation document corresponding to the original document.

14. An electronic device, comprising: at least one processor and memory;

the memory stores computer execution instructions;

the at least one processor executing the computer-executable instructions stored by the memory causes the at least one processor to perform the document translation method of any of claims 1-12.

15. A computer-readable storage medium having computer-executable instructions stored thereon which, when executed by a processor, implement the document translation method of any one of claims 1-12.

16. A computer program product comprising computer executable instructions which, when executed by a processor, implement a document translation method according to any one of claims 1 to 12.

Images