JPH02278368A - Machine translation input processing method - Google Patents

Abstract

PURPOSE:To improve the efficiency of translating processing by generating a sentence conforming to false conception structure by using the determined expression of the end of sentence and a word to be inputted to the node of the false conception structure so as to express the contents of an original sentence, and making this sentence the input of the translating processing. CONSTITUTION:An end of sentence and predicative part determining part 1 which is a means to determine the expression of the end of sentence from prescribed sentence end information and a predicate to be inputted in conformity to the contents of the original sentence, a false conception structure determining part 2 which is the means to determine and display the prescribed false conception structure to show the constitution of the sentence to be determined from this predicate, and a sentence generating part 3 which is the means to generate the sentence conforming to the false conception structure by using the determined expression of the end of sentence and the word to be inputted to the node of the false conception structure so as to express the contents of the original sentence are provided, and the generated sentence is made the input of the translating processing. By making a user input a key word to express the required contents of the original sentence, the sentence of clear syntax is generated automatically. Thus, the efficiency of the translating processing is improved, and the rate of a translation error is reduced.

Description

[Detailed description of the invention] [Summary] Regarding the preprocessing of original text generation in computer machine translation, by inputting a keyword indicating the required content, the original text with a syntax suitable for machine translation processing is automatically generated. , aiming at a machine translation input processing method that streamlines translation processing, when translating an original text in a given language into a text in another predefined language.
a means for determining a sentence-final expression from predetermined sentence-final information and a predicate input according to the content of the original sentence; a means for determining and displaying a predetermined pseudo-concept structure representing a sentence structure determined from the cough predicate; A means is provided for generating a sentence according to the pseudo-conceptual structure using a sentence-final expression and a word inputted to a node of the pseudo-conceptual structure to express the content of the original sentence, and the generated sentence is subjected to the translation process. TECHNICAL FIELD The present invention relates to machine translation processing in a computer, and a machine translation input processing method for automatically generating an original text so as to improve the efficiency of the translation processing.

[Prior art and problems to be solved by the invention] As is generally known in machine translation processing, by referring to a word dictionary, the original text is divided into words, numbered, and parts of speech are determined, and their syntactical parts are determined. As a result of analyzing the positioning according to the I-syntax rules and performing so-called semantic analysis, a conceptual structure indicating the conceptual structure of the sentence is generated using the predicate as the central node and the relationships connecting the syntactic element nodes. When the conceptual structure is correctly generated, based on it, select the words of the translation language that correspond to each word in the dictionary, and connect them according to the syntactic rules of the previous sentence to create a correct translation. can generate sentences.

In this case, the original text used is one freely written by a person, so it is easy to create complex sentences during syntactic analysis, and ambiguous sentences can result in incorrect syntactic analysis results, leading to mistranslations. It is also easy to reach.

Therefore, in order to perform translation processing efficiently, it is possible to input an original text with clear syntax, but it is difficult to restrict the original texts created by various users of a translation system in this way.

The present invention aims to provide a machine translation input processing method that can improve the efficiency of translation processing and reduce the rate of mistranslation by automatically generating an original text with a syntax suitable for machine translation processing by inputting a keyword indicating the desired content. do.

[Means to solve the problem] FIG. 1 is a block diagram showing the configuration of the present invention.

The figure shows the configuration of a machine translation input processing method, in which, when an original sentence in a predetermined language is translated into a sentence in another predetermined language, a means for determining a sentence-final expression from predetermined sentence-final information and predicates input according to the content of the original text. a sentence-final/predicate determining unit 1, which is a means for determining and displaying a predetermined pseudo-concept structure indicating the structure of a sentence determined from the predicate; A sentence generation unit 3 is provided as a means for generating a sentence according to the pseudo-conceptual structure using words input to nodes of the conceptual structure to express the content of the original sentence, and the generated sentence is subjected to the translation process. As input.

[For production]

With the above processing method, a sentence with clear syntax is automatically generated by the user inputting a keyword that represents the content of the original text, which improves the efficiency of the translation process and reduces the rate of mistranslations. reduction is possible.

〔Example〕

When this system shown in FIG. 1 is started, the processing unit 10 of the sentence ending/predicate determination unit 1 first inputs sentence ending information.
For example, a list of specified items is displayed on the user's terminal 11 and the user is prompted to select desired information.

The information at the end of the sentence is about the predicate, for example, U explanation.” [Planned]
``Possible'' ``Type of content of sentences such as request J, ``affirmative'' [Negative J, ``declarative'' and ``interrogative'' sentences, past
Four tenses are specified, such as (did), present (to do), etc., and the user selects, for example, ``planned, affirmative, declarative, and to do.''

The processing unit 10 has a verb dictionary file 12 and a verb conjugation list file 13, and then displays a list of verbs in the verb dictionary file 12 on the terminal 11 to allow the user to select a verb for the predicate.

For example, if the user selects the verb "to make a phone call," the processing unit 10 uses the verb dictionary file 12 to
For example, read the verb dictionary record for the verb "to call" shown in Figure 2, and its conjugation number is "1", so
From the verb conjugation list file 13, for example, conjugation number r1. shown in FIG.・Use - Read out the list.

Since ``to make a phone call'' is in the su variable system, this is a list of su flexural forms, and from this list, the processing unit 10 selects ``scheduled, affirmative, ~suru j'' in the previously specified sentence ending information. Refer to the column that corresponds to , and determine [I intend to do so. is used as a sentence-final expression.

The processing unit 10 passes the above processing results to the pseudo concept structure determining unit 2. The processing unit 14 of the IM similar concept structure determining unit 2 is a PCS.
It has file 15 and node file 16, and first
Conceptual structure number r5 of the verb dictionary record for "to call" shown in the figure. For example, the conceptual structure number "5" shown in FIG.
5J, the node record with the number specified in the 5 columns from the 1st node to the 57th node is searched and read from the node file 16.

Each node record in the node file 16 is, for example, the fifth
The configuration is as shown in each column of the figure, and! The processing unit 14 before pseudo-concept structure determination 2 determines the display position j of each node record.
By displaying the explanatory word specified in the "Node description" column as a node at the position on the display screen of the terminal 11 specified by the X and Y coordinates of the column, and relating them by connecting them, for example, with a line, the translation can be performed. For example, as a pseudo-concept fI structure that simplifies the conceptual structure of the analysis result of the original text described in the processing explanation,
The process is completed by displaying as shown in the figure, and the process result is passed to the text generation section 3.

The user can then fill in the required blank space (indicated by a rectangular frame in the figure) of each node of the displayed one pseudo-concept structure with the user's intended information, such as ``Me, after - hours.'' Enter the word that describes the content.

When the sentence generation unit 3 reads them, it arranges the words in the order determined for each conceptual structure, for example, and inserts the particle specified in the "Connecting Particle J column" of the node record in FIG. By adding the sentence-final expression determined by the sentence-final/predicate determination unit 1 to the stem of the predicate, the sentence ``I'm going to call you in an hour'' is generated, and this sentence is input into the translation processing system. For example, start the process of translating into English.

As mentioned above, users can specify end-of-sentence information such as ``schedule'', affirmation, and ``J'', and use the keyword ``call''.
By inputting ``I J [-hours later J'' according to the system's guidance, a sentence such as ``I'm going to call you in an hourJ'' will be automatically generated, and the sentence will have a text that is easy to translate. The system can adopt various syntaxes.

〔Effect of the invention〕

As is clear from the above description, according to the present invention, in machine translation using a computer, by inputting a keyword indicating the desired content, an original sentence with a syntax suitable for machine translation processing is automatically generated and input for translation processing. Therefore, there is a significant industrial effect of increasing the efficiency of translation processing and reducing the rate of mistranslation.

Figure 5 is an explanatory diagram of node records, FIG. 6 is a diagram showing a display example of a pseudo-concept structure.

In the figure, 1 is the sentence ending/predicate determining part, 2 is the conceptual structure determining part,
3 is a sentence generation unit, l0114 is a processing unit, 11 is a terminal, 12 is a verb dictionary file, 13 is a verb conjugation list file, 15 is a PCS file, 16 is a node file

[Brief explanation of drawings]

Figure 1 is a block diagram showing the configuration of the present invention, Figure 2 is an explanatory diagram of a verb dictionary record, Figure 3 is an explanatory diagram of a verb conjugation list, and Figure 4 is an explanatory diagram of a pseudo-concept structure record. Bronch diagram showing the structure of the invention Diagram explanatory diagram of pseudo-conceptual structure record Diagram verb conjugation - explanatory diagram Figure 3 Diagram explanatory diagram of node record Figure 5

Claims (1)

[Claims] Means (1) for determining a sentence-final expression from predetermined sentence-final information and predicates input according to the content of the original text when translating an original sentence in a predetermined language into a sentence in another predetermined language; means (2) for determining and displaying a predetermined pseudo-conceptual structure indicating the structure of a sentence determined from a predicate; and inputting the determined sentence-final expression and the node of the pseudo-conceptual structure to express the content of the original sentence. A machine translation input process characterized in that it is configured to include means (3) for generating a sentence according to the pseudo-conceptual structure using words, and to use the generated sentence as input to the translation process. method.