JPH0454671A - Natural language translating device - Google Patents

Abstract

PURPOSE:To efficiently execute the processing by providing a means for replacing an input sentence with a different expression executed by first natural language, based on a result of analysis, converting the input sentence whose literal translation is difficult to a sentence of an equivalent expression whose literal translation can be executed, and thereafter, converting it to natural language. CONSTITUTION:A Japanese sentence inputted from an input device is subjected to word division and syntax analysis by an analyzing device 2, and with respect to this result, by a rewriting device 3, the above word and syntax information are replaced, based on a rewriting table 8. Subsequently, in a converter 4, the word and the syntax information are converted from Japanese to English by using a conversion dictionary 9, and in a generating device 5, an English sentence is generated by using a generation dictionary 10, and thereafter, it is outputted from an output device 6. In such a way, a sentence of original language whose literal translation is difficult is formed so that the mechanical processing can be executed, and the preliminary processing is executed, therefore, a sentence of original language whose literal translation is difficult can also be processed efficiently.

Description

Detailed Description of the Invention [Field of Industrial Application] The present invention is a natural language translation device, that is, a sentence in a first natural language (original language) read from an input device, into a second natural language (original language). The present invention relates to a device for automatically translating and outputting sentences in a target language, and particularly to a natural language translation device that can efficiently process sentences in a source language that are difficult to translate directly.

[Conventional technology]

As natural language translation devices, many devices based on a so-called transfer type are known.

This method analyzes sentences in the source language, generates an intermediate representation in the source language based on the results, converts (transfers) this intermediate representation into an intermediate representation in the target language, and then It consists of processing steps that ultimately generate sentences in the target language. Regarding these, for example, the description in "Machine Translation System (MU Project) J (Information Processing Society of Japan Natural Language Processing Study Group Material 84-NL47-10 (1985))" is helpful.

[Problem to be solved by the invention]

In the one-way transfer method described above, the intermediate representation of the source language generated after analyzing the input sentence of the source language basically has a format that conforms to the structure of the input sentence. For this reason,
If the input sentence in the source language is an expression that can be directly translated, the output sentence in the target language will be relatively well-formed. In such a case, there is a problem that the resulting output sentence in the target language does not make any sense as it is. Conventionally, the source language,
An operator familiar with both target languages would monitor the output sentences in the target language and, if necessary, take action by rewriting the input in the source language.

In order to avoid the above-mentioned problems, at the stage of analyzing input sentences in the source language, expressions in the source language that are difficult to translate are fixed and used as idiomatic expressions in the target language prepared in advance. There is such a way to uniquely convert it, and it has actually been done as needed. This means, for example, that the Japanese “
It is like making "based on" correspond to the English word "based on-". However, depending on the context, this method cannot deal with expressions that should not be regarded in a fixed manner. At the same time, due to the influence of a fixed understanding, the analysis of other expressions does not go well, and the problem often arises that the translation fails. For example, in the case of ``Company A's translation device is based on a single transfer system,'' if ``based on'' is taken fixedly, the analysis of other parts may not work. It is.

The present invention has been made in view of the above circumstances, and its purpose is to solve the above-mentioned problems in the conventional technology, and to enable natural language translation that can efficiently process sentences in the source language that are difficult to translate directly. The goal is to provide equipment.

[1) Means for Solving II] The above-mentioned object of the present invention is to provide an input means for reading sentences in a first natural language, and to perform word segmentation and syntax analysis on the sentences read by the input means. an analysis means, a conversion means for converting the input sentence into a second natural language based on the result of the analysis by the analysis means, and a conversion result by the conversion means to be arranged as a natural sentence in the second natural language. In a natural language translation device having a generation means, means is provided for replacing the input sentence with another expression in the first natural language based on the result of the analysis by the analysis means, and direct translation in the first natural language is provided. This is achieved by a natural language translation device that converts a difficult input sentence into a sentence with an equivalent expression that can be directly translated, and then converts it into the second natural language.

[Effect]

In the natural language translation device according to the present invention, the input sentence is
We have provided a means to substitute another expression in the source language based on the results of the analysis by the analysis means, so that we can perform preliminary processing to convert sentences in the source language that are difficult to translate directly into a machine-processable form. , it becomes possible to efficiently process sentences in the source language that are difficult to translate directly.

〔Example〕

Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a diagram showing the basic block configuration of a natural language translation device that is an embodiment of the present invention. In the figure, l is an input device for inputting a sentence in the source language, 2 is an analysis device that divides the input source language sentence into words and parses it by referring to an analysis dictionary 7, and 3 is described in detail later. A rewriting device 4 is a conversion device that converts the input sentence information rewritten by the rewriting device 3 into target language information using a conversion dictionary 9;
Reference numeral 5 denotes a generation device that generates sentences in the target language using the generation dictionary IO, and 6 denotes an output device that outputs the generated sentences in the target language.

An outline of the operation of the natural language translation device shown in this embodiment is as follows. In the following explanation, the source language is Japanese and the target language is English.

The Japanese sentence input from the input device 1 is word-divided and syntactically analyzed by the analysis device 2, and the above-mentioned words and syntactic information are replaced by the rewriting device 3 based on the rewriting table 8. be done. Next, the conversion device 4 uses the conversion dictionary 9 to convert the words and syntactic information from Japanese to English, and the generation device 5 uses the generation dictionary lO to generate an English sentence, which is then output. It is output from the device 6.

FIG. 2 is a block diagram showing the processing contents of the above-mentioned rewriting device 3, in which the key word detection unit 31. Pattern matching section 32.
It is composed of a word syntax information generation section 33. Further, FIG. 3 is a diagram showing a configuration example of a rewriting rule table 8 that controls the operation of the rewriting device 3. As shown in FIG.

The operation of the natural language translation apparatus of this embodiment configured as described above will be explained in detail below using the input sentence examples shown in FIGS. 4(a) and 4(b).

First, in the case of an input sentence a shown in FIG. 4(a), the input sentence a is divided in the analysis device 2 as shown in the word division result a, and is parsed as shown in the syntactic analysis result a. Note that the syntax analysis results are generally shown in Figure 4 (c), for example.
), but here, for convenience, the above-mentioned syntax analysis result a will be shown.

In the rewriting device 3 that has received the above analysis result, the key word detection unit 31 first examines the control rule table 8 and obtains the rule 51 for "based on". Next, the pattern matching section 32 matches the pattern condition specification of the above-described rule 51 with the word segmentation results and the syntax analysis results. That is, first, X2 of the rule 51 including "based on" and clause 4 are associated with each other, and it is determined whether the structures match.

In this case, they match, so next, Xi and clause 3 related to clause 4 are associated, and X3 and clause 5 are associated, and it is determined whether the structures of each match. In this case, since they all match, the condition is specified as ro KJ and the matching is completed.

Finally, the word syntax information generating unit 33 generates the word "based on" and changes the word structure of the bunsetsu 3 in accordance with the change of the word syntax information in the rule 51.

By deleting clause 4, the word division result and the syntactic analysis result are word division result a' shown in FIG. 4(a).

Change the syntax analysis result as shown in a'.

In the case of the input sentence shown in FIG. 4(b), as in the case of the input sentence a shown in FIG. The results are analyzed as shown below. Furthermore, in the rewriting device 3, the key word detection unit 31 detects the rule 5 for “based on”.
1 is obtained, the process proceeds to the pattern matching section 32, and the above-mentioned rule 5 is executed.
Compare the pattern condition specification in step 1 with the word segmentation results and syntax analysis results. In this case, the determination result of whether or not the structure of associating x2 of rule 51 that includes "based on" and phrase 4 matches is a mismatch, so this rule is not applied. Therefore, in this case, rewriting as in the case of FIG. 4(a) is not performed.

By performing the processing described above, in the case of the input sentence a in Figure 4(a), the particle equivalent word "based on" is recognized, and in the case of the input sentence in Figure 4(b), the particle equivalent word "based on" is recognized. ” can be left as a verb.

In addition, as for the processing after the particle equivalent word "based on" is generated, the same as the conversion dictionary 9 used in the conversion device 4 shown in FIG. All you need to do is register the information shown in . As a result, for each clause shown in FIG. 4(a), clause 1: of the company A clause 2: the natural language translation device is -11
ranslation system -4 based on clause 3
On Transfer Method Clause 4: After a transformation such as * be designed, the following translated sentence will be generated.

The natural language tran
slation system of the comp
any was designed based
onTransfer Method.

Incidentally, if a translated sentence is generated from the word division result a and the syntactic analysis result a shown in FIG. 4 as in the conventional method without performing such rewriting, the following translation will be difficult to understand.

Clause 1: -4 of the company A Clause 2: The natural language translation device is one* the natural language t
ranslationystem Clause 3 Nidoran Shua one way → to Transfer Method Clause 4: Based on one* and --base Clause 4: Designed → be designed The natural language tran
slation system of the con+
Pany was designed and?
? ? base to transfer method
.

here,? ? ? indicates that the subject, which is required in English, was not found in the analysis.

The above-mentioned embodiment shows an example of the present invention, and it goes without saying that the present invention should not be limited thereto. For example, in the above embodiment, the rewriting device 3 has been described as having the control rule table 8 fixedly installed in the device, but the control rule table 8 is stored in a freely replaceable storage medium. It is also possible to store the information and use the most suitable one depending on the field of use, for example.

[Effects of the Invention J As described in detail above, according to the present invention, the input means for reading a sentence in a first natural language and the word segmentation for the sentence read by the input means.

an analysis means for performing syntax analysis; a conversion means for converting the input sentence into a second natural language based on the result of the analysis by the analysis means; In the natural language translation device, there is provided a means for replacing the input sentence with another expression in the first natural language based on the result of analysis by the analysis means, Since an input sentence in a natural language that is difficult to directly translate is converted into a sentence with an equivalent expression that can be directly translated, and then converted to the second natural language, it is possible to efficiently process sentences in the source language that are difficult to directly translate. This has the remarkable effect of realizing a possible natural language translation device.

[Brief explanation of drawings]

Fig. 1 is a diagram showing the basic block configuration of a natural language translation device that is an embodiment of the present invention, Fig. 2 is a block diagram showing the processing contents of the rewriting device, and Fig. 3 is a diagram showing the operation of the rewriting device. FIG. 4 shows an example of the configuration of a rewriting rule table.
The figure shows input sentences and their analysis results to explain the operation of the rewriting device. Figure 5 shows the expressions that should be registered in the conversion dictionary in order to convert the derived expressions in the rewriting device into the target language. FIG. 3 is a diagram showing an example of information. I: input device, 2: analysis device, 3: rewriting device, 4:
Conversion device, 5: Generation device, 6: Output device, 9; Conversion dictionary, 102 Generation dictionary. 7: Dictionary for analysis, 8: Rule table for rewriting, Figure Figure Word information/Syntax information Word information/Syntax information Figure (Part 3) (c) The designed/ natural translation device is Transfer, one way

Claims (2)

[Claims] (1) An input means for reading a first natural language sentence (input sentence), an analysis means for word segmentation and syntactic analysis of the sentence read by the input means, and an analysis means for analyzing the result of the analysis by the analysis means. In the natural language translation device, the natural language translation device includes a conversion means for converting the input sentence into a second natural language, and a generation means for preparing the conversion result by the conversion means as a natural sentence of the second natural language. Means is provided for replacing the input sentence with another expression in the first natural language based on the result of the analysis by the means, and the input sentence in the first natural language that is difficult to be directly translated is replaced with an equivalent expression that can be directly translated. A natural language translation device characterized in that the natural language translation device converts the sentence into the second natural language. (2) The means for replacing the input sentence with another expression in the first natural language is conditioned on one or more of word information, word combinations, and syntactic information that are the results of analysis by the analysis means. . The natural language translation device according to claim 1, wherein the natural language translation device is controlled by an expression substitution rule table having an equivalent expression pattern corresponding to the expression substitution rule table.