RU2667030C1 - System and method of intellectual automatic selection of perfomers of translation - Google Patents

Abstract

FIELD: information technology.SUBSTANCE: invention relates to a system and method for intelligent automatic selection of translators. System consists of a cloud-based network server, a user interface, a translation memory base, a common glossary module, individual modules of glossaries for each translator, a module for automatic matching of terminology translated by performers of earlier texts with the terminology of the source file for translation, module calculation for each performer translating the numerical characteristics of the proximity of the terminology of the source file and the terminology of the translator, module for automatic assessment and forecasting of the quality of the translation of the source file by the translators, taking into account the subject matter of the document, the module for automatically calculating the speed of each performer, the module for automatic maintenance of the calendar of translators, module for automatic calculation of the cost of translation of the source file and selection of performers, taking into account the specified rate in the profile, the module for automatically optimizing the selection of the most suitable translators based on the criteria listed above.EFFECT: technical result consists in automating the selection of performers for translation.8 cl, 5 dwg

Description

FIELD OF THE INVENTION

The present invention describes a system and method for intelligent automatic selection of translators.

BACKGROUND

The collection and exchange of information for any scientific, commercial, political or social purpose often requires a quick and effective translation of the text so that a lot of knowledge and ideas become useful on a global scale. Computer programs that automatically translate from one language to another ("machine translation programs") can, in principle, satisfy this need, and such programs have been developed and continue to be developed for many languages. For a formal presentation style in deeply studied languages (as opposed to an informal, idiomatic or colloquial style), such machine translation programs demonstrate a fairly adequate translation quality. For a less formal presentation style, the quality of machine translation is still insufficient for a full understanding of the meaning of the texts. For more difficult or less studied languages (e.g., Arabic), existing machine translation programs do not work well even for formal communication (e.g., Modern Standard Arabic), and they are especially weak in the case of informal, conversational, and idiomatic communication.

Similarly, where high-quality accurate translation is required, machine translation alone is not enough even for well-studied languages (for example, English, French, Spanish, German and other languages).

Professional translators are able to provide high-quality translations for difficult languages and informal communications, but Internet applications require constant availability and prompt response, which cannot be guaranteed if existing approaches to organizing the work of professional translators are used. The prior art application US 2015/0120273 (A1) "NETWORKED LANGUAGE TRANSLATION SYSTEM AND METHOD", publ. 04/30/2015. This application describes the technology of automated work of translators in a distributed system, in which the source file is translated by many translators connected to the cloud server, the lexical, morphological and syntactic analysis of logical segments into which the source file is divided is described, and translation memory is used (translation memory ) and a glossary module for storing translation terms.

A significant drawback of this technical solution is that it does not provide the possibility of fully automatic selection of performers for translation.

SUMMARY OF THE INVENTION

This technical solution is aimed at eliminating the disadvantages inherent in existing solutions from the prior art. This technical solution is aimed at expanding the arsenal of technical means of a specific purpose, in our case, expanding the arsenal of technical means of intelligent automatic selection of translators, and as a technical result achieved by the claimed solution, the claimed invention may implement the specified purpose, namely, the implementation of intelligent automatic selection of translators.

This technical result is achieved thanks to a system of intelligent automatic selection of translators, consisting of:

a cloud network server that provides access for many performers and many translation customers connecting to the server via the Internet;

a user interface that allows many translation customers to upload source files for translation into the system of intelligent automatic selection of translation performers and receive information about proposed translation performers;

translation memory database storing previously translated texts of all artists in the system with key terms highlighted in these texts; the general glossary module, made in the form of numbered clusters, each of which contains key terms highlighted in the translated texts that are similar in meaning and, accordingly, related to the topic, with each term being assigned the cluster number to which it belongs;

individual glossary modules for each translator, indicating the specific topic in which one or another artist works; a module for automatically matching the terminology of texts previously translated by the performers with the terminology of the source file for translation, while for terms extracted from the source file for translation, it is calculated how many times each term appears in it, thus, the terminological frequency vector is calculated;

a calculation module for each translation executor of a numerical characteristic of the closeness of the terminology of the source file and the terminology of the translation executor and ranking the translation executors by this numerical characteristic;

a module for automatic assessment and forecast of the quality of the translation of the source file by the translators, taking into account the subject of the document;

a module for automatically calculating the speed of each artist performing the calculation of the number of words translated per hour; a module for automatically maintaining the calendar of work of translation artists, which allows real-time analysis of the availability of a translation artist for work, taking into account the factor of its employment when searching for a translation artist;

a module for automatically calculating the cost of translating the source file and selecting artists based on the rate specified by them in the profile;

module for automatic optimization of the selection of the most suitable translators based on the above criteria.

This technical result is also achieved thanks to the method of intelligent automatic selection of translators, consisting of the following steps:

receiving a request to translate the source file in a distributed network system;

extraction of key terminology from the source file;

search for similar terminology in the general glossary, consisting of numbered clusters, each of which contains key terms highlighted in the translated texts that are similar in meaning and, accordingly, related in topic, with each term being assigned the cluster number to which it belongs; and

search for similar terminology in the individual glossaries of each translator, which indicate in which specific topic a particular artist works;

comparing the terminology of the performers with the terminology of the source file for translation, while for terms extracted from the source file for translation, it is calculated how many times each term appears in it, thus, the terminological vector of frequencies is calculated; based on this comparison, the selection of artists working in the subject of the source file;

automatic analysis of the quality of previously translated texts of selected artists;

automatic calculation of the speed of each selected artist;

real-time analysis of the availability of selected artists for translation;

automatic calculation of the cost of translating the source file for each selected artist; and

based on the above steps, the automatic selection of the most suitable translators.

In some embodiments, the implementation of the technical solution additionally carry out the selection of performers for a language pair.

In some embodiments of the technical solution, the system further comprises a module for syntactic, morphological and linguistic text filtering.

In some embodiments of the technical solution, the glossaries are replenished with new terms during the work of the performers as they accumulate new translated texts.

BRIEF DESCRIPTION OF THE DRAWINGS

The implementation of the invention will be described hereinafter in accordance with the accompanying drawings, which are presented to illustrate the essence of the invention and in no way limit the scope of the invention. The following drawings are attached to the application:

FIG. 1 - depicts the search for artists working in the subject of the source document for translation;

FIG. 2 - depicts the task of selecting related texts on the subject to take into account the specifics of translation tasks;

FIG. 3 - displays the construction of mathematical models for constructing estimates and forecasting the quality of translations;

FIG. 4 - displays the automatic selection of translators matching the specified criteria.

FIG. 5 - presents a General view of the system of intelligent automatic selection of translators that implements the claimed method.

DETAILED DESCRIPTION OF THE INVENTION

Translation is an intellectual activity that requires highly qualified performers, for its high-quality implementation, the translator needs to understand the text in the original language and be able to adequately formulate it in the target language in accordance with the norms of the target language. In the practice of real work, this means that translators are an extremely heterogeneous resource, for example, a translator who performs high-quality literary translations is most likely not able to translate a technical or medical text.

The selection of performers and the organization of work on the translation project is a topical issue in the linguistic services market. For all possible topics and language pairs, not a single employee of the company will be able to evaluate the translation.

In the following detailed description of the implementation of the invention, numerous implementation details are provided to provide a clear understanding of the present invention. However, to a person skilled in the art, it will be apparent how the present invention can be used, both with and without implementation details. In other instances, well-known methods, procedures, and components have not been described in detail so as not to obscure the understanding of the features of the present invention. In addition, from the foregoing it will be clear that the invention is not limited to the above implementation. Numerous possible modifications, changes, variations and replacements preserving the essence and form of the present invention will be apparent to those skilled in the subject field.

The present invention is directed to providing a system and method for intelligent automatic selection of translators. The claimed system and method of intelligent automatic selection of translators use a cloud network server and a distributed network language translation system - this is a distributed network of professional translators and professional translation tools (translation memory, machine translation systems, etc.) that interact through the program and user interfaces of the system and together they perform real-time translation of texts for which the use of exclusively typewriter is not enough translation or traditionally organized professional translation, including translation of dynamic communications and other texts created in various information environments.

The distributed network language translation system provides tools for aggregating the resources of a large number of translators, with different access modes, with various professional skills, both professional translators and computer-based machine translation systems, for efficiently performing high-quality translations in real time.

To search for translators, from which the optimal translators for the project will subsequently be selected, an algorithm is used consisting of the following steps:

The selection of key terminology from all texts of all translators in the system and the compilation of a common module of glossaries based on this terminology and individual glossaries for each translator, indicating the specific subject in which one or another artist works. This is necessary to reduce the amount of data analyzed, as well as to develop a selection criterion for a translator based on his ability to work with specific terminology.

Partitioning the extracted terminology of all performers into classes. Classes are determined automatically by clustering using machine learning.

An example of the implementation of such a clustering algorithm, you can use the K-means algorithm (K-means algorithm). This algorithm works on the principle of maximizing the distance between the final clusters. This algorithm belongs to the class of quadratic error algorithms, since the ultimate goal of the algorithm is to minimize the mean square error of the partition calculated by the formula:

where X is the aggregate of data to be divided into clusters,

K is the number of clusters,

n _j - number of elements belonging to a cluster with index j,

- i-тый элемент кластера с номером j,

- i-th element of the cluster with number j,

|| a-b || distance from element a to element b in metric space.

The objective of this algorithm is to minimize the sum of squared distances from all elements to be clustered to the centers of the clusters to which they belong. After applying clustering, the resulting clusters are numbered and each term from the glossary is assigned the number of the cluster to which it belongs. Each individual cluster is hereinafter referred to as generalized topics (quasithematics). All corpses of texts of all performers are assigned vector themes. To do this, for each text of translators from the general system case, the frequencies of terms falling into one or another cluster are calculated. Thus, each text is presented as a vector with a dimension equal to the number of clusters (a point in a multidimensional space), and it is known about which translator it belongs to. In a similar way, the vector of topics of the new text for translation is calculated.

On the basis of a detailed comparison of the terminology of texts previously translated by translators, this terminology is compared with the terminology of the file to be translated. For the terms extracted from the source text for translation, it is calculated how many times each term appears in it, thus, a terminological frequency vector ( a _i , ..., a _k ) is obtained.

For each translator, a numerical characteristic of the proximity of the text and the artist is calculated:

where w _i - the frequency of this term in the texts of the selected artist (T).

Further, those translators whose thematic vectors have topics that are far from the topics of the source file that needs to be translated are excluded from consideration, which allows to narrow the space of translators and reduce the computational load on the system.

Next, the translators are ranked by this numerical characteristic and then at the next stages pass an additional selection based on restrictions on quality, cost and terms set by the client.

After that, the system proceeds to the automatic assessment and prediction of the quality of the translation of the source file by the translators. The following information is used for this:

- Analysis of low-level data on translation segments - time spent in the segment, number of actions of the executor, number and type of edits between the stages of work (for example, editing the editor after the translator, editing the client after the editor), pressing certain keys, using interface elements, service shortcut keys .

- Analysis of the compliance of the translation of the terminology of the document with the glossary of the project or automatically created terminology at the stage of analysis of topics; if the terms are present in the project glossary, but were incorrectly translated in the text, this is classified as a serious mistake. If some terms are not present in the glossary, but are descriptive, then it is analyzed how these terms are translated in other documents falling into the same thematic cluster as this document. If within the cluster these terms are translated quite uniformly, i.e. There is a statistically preferred translation option (one or several) for this topic, and the translation of the term in the document does not coincide with it, the error is classified as serious.

- The results of automatic checks: spelling, grammar, punctuation, preserving the structure and order of segment tags, saving placeholders, the presence of extra and double spaces, controlling context matches, keeping dates and other numerical parameters, keeping the case of words, having the same translations for different source segments, and vice versa - different translations for the same source segments, erroneous repetitions of words, a combination of Latin and Cyrillic letters in one word.

- A set of various linguistic descriptors (such as the average length of a sentence, a wealth of vocabulary, text complexity, etc.)

- Expert assessments (both integral according to a single quality index, and detailed, with classification of errors).

To build an automatic quality assessment, various machine learning models are used. For this:

- A correlation model is being built between automatically measured parameters and edits (their volume and essence) at the editing stage.

- A correlation model is being built between edits at the editing stage and the human (manual assessment) translation quality.

- Validation and refinement of the constructed models are carried out on fixed test tasks and the results of checks according to the formal procedure with typification of errors.

- The final model of the correlation between automatically measured parameters and the final human assessment is under construction.

Further, the constructed mathematical model is used to assess and predict the quality of all translations performed, including those in which there are no corrections at the editing stage. The model is iteratively rebuilt with the accumulation of new data (estimates, corrections, etc.) to increase the accuracy of the algorithm.

An important factor in the selection of performers is their willingness to complete the task. The most sought-after performers, as a rule, are more loaded with work. In the product, all information about each artist is available for analysis - current and planned projects, data on the average speed of work. This allows you to create an automatic calendar of work of the contractor and additionally take into account the factor of his employment during the search. Given the employment factor of translators when building the translation process, the system helps to find hidden reserves, allowing the translator to increase his earnings.

In addition, speed is an important factor in selecting artists, and the speed of the artist’s work is determined automatically based on an analysis of the number of words translated per hour.

Also, using the module for automatic calculation of the cost of the translation of the source file, artists are selected taking into account the rate indicated by them in the profile, if this rate falls under the amount indicated by the customer that he is willing to spend on the transfer.

Based on the above steps and selection criteria, the most suitable translators of the source file are automatically selected.

After that, the system sends out notifications of the invitation to participate in the translation project to the selected performers, each respective artist confirms or rejects this invitation. After confirmation of participation, the designated contractor enters the system and, through the web-based user interface, starts work on the translation of the document.

The translated document is then automatically transmitted to the translation customer via the web user interface from which the customer downloads the translated file. If the source files came from a certain external information system, then the translated files can be placed into the same system using the program APIs of the interfaces contained in the integration layer.

In addition, it is possible to additionally apply syntactic and morphological filtering of the text, for example, cleaning text from meta-information, tags and formatting, marking up with parts of speech, converting words to the main form), as well as linguistic filtering.

The present invention in its various embodiments is implemented as a system and method implemented on a computer.

In FIG. 5 shows a General view of the system (100) that implements the claimed method.

The system of intelligent automatic selection of translation executors (100) can be performed on the basis of a wide range of electronic computing devices, for example, a personal computer, laptop, server cluster, etc.

In the General case, the system (100) contains one or more processors (101) that perform the main computing work when implementing the steps of the method.

Random access memory (RAM) (102), designed for online storage of instructions executed by one or more processors (101). The storage medium (103) can be a hard disk (HDD), solid-state drive (SSD), flash memory (NAND-flash, EEPROM, Secure Digital, etc.), an optical disk (CD, DVD, Blue Ray) , mini disk or their combination.

Input / output (I / O) interfaces (104) are standard ports and devices for pairing devices and transmitting data, selected based on the required configuration of the system (100), in particular: USB (2.0, 3.0, USB-C, micro, mini), Ethernet, PCI, AGP, COM, LPT, PS / 2, SATA, Fire Wire, Lightning, etc.

I / O facilities (105) are also selected from a well-known range of different devices, for example, a keyboard, touchpad, touch display, monitor, projector, mouse, joystick, trackball, light pen, stylus, sound output devices (speakers, headphones, built-in speakers buzzer) etc.

Data transmission means (106) are selected from devices designed to implement the communication process between different devices via wired and / or wireless communication, in particular, such devices can be: GSM modem, Wi-Fi transceiver, Bluetooth or BLE module, GPS module, Glonass module, NFC, Ethernet module, etc. System components (100) are interfaced via a common data bus (110).

Modifications and improvements to the above-described embodiments of the present technology will be apparent to those skilled in the art. The preceding description is provided as an example only and is not subject to any restrictions. Thus, the scope of the present technology is limited only by the scope of the attached claims.

Claims (28)

1. The system of intelligent automatic selection of translators, consisting of: a cloud network server that provides access for many artists and many translation customers connecting to the server via the Internet, and interacts with all parts of the intelligent automatic selection of translation artists; a user interface that allows many translation customers to upload source files for translation into the system of intelligent automatic selection of translation performers and receive information about proposed translation performers; translation memory database storing previously translated texts of all translators in the system with key terms highlighted in these texts; the general glossary module, made in the form of numbered clusters, each of which contains key terms highlighted in translated texts that are similar in meaning and correspondingly related in topic, with each term being assigned the cluster number to which it belongs; individual glossary modules for each translator, indicating the specific topic in which one or another artist works; a module for automatically comparing the terminology of texts previously translated by the performers with the terminology of the source file for translation, while for the terms extracted from the source file for translation, it is calculated how many times each term appears in it, thus the terminological frequency vector is calculated; a calculation module for each translation executor of a numerical characteristic of the closeness of the terminology of the source file and the terminology of the translation executor and ranking the translation executors by this numerical characteristic, excluding translators whose thematic vectors have topics that are far from the source file; a module for automatic assessment and forecast of the quality of the translation of the source file by the translation executors, taking into account the subject of the document and the criterion of the quality of work specified by the client; a module for automatically calculating the speed of each artist performing the calculation of the number of words translated per hour; a module for automatically maintaining the calendar of work of translation executives, which allows real-time analysis of the availability of a translation executor for work, taking into account the factor of its employment when searching for a translation executor and a criterion of the cost of the work specified by the client; a module for automatic calculation of the cost of the translation of the source file and the selection of artists, taking into account the rate indicated by them in the profile and the criterion of the cost of the work specified by the client; module for automatic optimization of the selection of the most suitable translators, taking into account the subject of the source file and the criteria set by the client for quality, cost and timing of the work. 2. The system according to claim 1, in which additionally the selection of performers is carried out on a language pair. 3. The system of claim 1, further comprising a module for syntactic, morphological and linguistic text filtering. 4. The system according to claim 1, in which the glossaries are replenished with new terms during the work of the performers as they accumulate new translated texts. 5. The method of intelligent automatic selection of translation performers, implemented by the system according to claim 1 of the formula, consisting of the following steps: receiving the source file for translation; extraction of key terminology from the source file; search for similar terminology in the general glossary, consisting of numbered clusters, each of which contains key terms highlighted in the translated texts that are similar in meaning and correspondingly related in topic, with each term being assigned the cluster number to which it belongs; and search for similar terminology in the individual glossaries of each translator, which indicate in which specific topic a particular artist works; comparing the terminology of the performers with the terminology of the source file for translation, while for terms extracted from the source file for translation, it is calculated how many times each term appears in it, thus the terminological vector of frequencies is calculated; on the basis of this comparison, the selection of artists working in the subject of the source file, with the exception of translators whose subject vectors have topics that are far from the subject of the source file; automatic analysis of the quality of previously translated texts of selected performers according to the criterion set by the client for the deadline for completing the work; automatic calculation of the speed of each selected contractor according to the criteria of the deadline set by the client; real-time analysis of the availability of selected artists for translation; automatic calculation of the cost of translating the source file for each selected artist according to the criterion set by the client for the cost of the work; and based on the above steps, the automatic selection of the most suitable translators. 6. The method according to p. 5, in which additionally select performers for a language pair. 7. The method of claim 5, further comprising syntactic, morphological, and linguistic filtering of the text. 8. The method according to claim 5, in which the glossaries are replenished with new terms during the work of the performers as they accumulate new translated texts.

Images