JPH03116272A - Mechanical translation system containing sentence pattern dictionary - Google Patents

Abstract

PURPOSE:To improve the processing efficiency for analysis of an input sentence by performing simultaneously the retrieval and the collation of character patterns during analysis of a morpheme. CONSTITUTION:A morpheme analysis part 3 is provided together with a word dictionary 4, a pattern dictionary 5, and a word list/pattern candidate list screening part 6. An input text inputted to the part 3 undergoes the segmentation of character strings and the reference to dictionaries. Both dictionaries 4 and 3 are simultaneously retrieved and collated with each other in a dictionary reference mode. Then a word list or a pattern candidate list is obtained via the part 6. Thus the processing efficiency is improved in a mechanical translation system.

Description

[Detailed description of the invention] [4! I Summary] Regarding machine translation systems that perform pattern matching with sentence patterns registered in advance by the system user, as well as morphological analysis using a word dictionary when analyzing input sentences in the source language, conventionally, word dictionaries were used. After morphological analysis, the notation of the input sentence is examined again, and the above sentence patterns are searched and matched.With the aim of solving the inefficient problem, predetermined words in the sentence patterns are used as search keys. and a means for searching the pattern dictionary at the same time as searching the word dictionary during morphological analysis when analyzing an input sentence in the source language. .

[Industrial Application Field] The present invention relates to a machine translation system that converts an input sentence in a source language into an output sentence in a target language, and in particular, when analyzing an input sentence in a source language, it can be used in conjunction with morphological analysis using a word dictionary. , pattern matching with sentence patterns registered in advance by the system user (
Related to a machine translation system that performs "pattern matching" (also called "pattern matching").

[Prior Art] Generally, a machine translation system is composed of an expert system consisting of a translation engine and a knowledge base.

The translation engine is the core part of an expert system that parses input sentences in the source language using analytical grammar knowledge and generates output sentences in the target language using generative grammar knowledge. It consists of a dictionary of words in the target language.

In such a translation system, it is possible to output correct translation results for input sentences with regular grammatical expressions, but it is possible to output correct translation results for expressions with relatively complex grammatical structures, such as phrases specific to specialized fields. It is difficult to obtain translation results.

Therefore, the system user can register in advance the input sentences and their parallel translations regarding phrasings specific to the above specialized fields and other idiomatic expressions in the sentence pattern table as necessary, and then analyze the input sentences. At this time, a method has been proposed in which, in addition to morphological analysis using a normal word dictionary, pattern matching is performed against the sentence pattern table.

This will be explained in more detail below.

FIG. 3 is a diagram illustrating a conventional translation processing method using a sentence pattern table, and shows only the analysis portion of an input sentence in the source language that is directly related to the present invention.

A human text, which is an input sentence in the source language, is broken down into individual parts of speech by cutting out characters from the input sentence and looking it up in a dictionary using a word dictionary 52 in a morphological analysis unit 51 to obtain a single IJ sentence.

Further, FIG. 4 is a diagram for explaining the concept of the above-mentioned Single Kogo dictionary, and morphological analysis is performed by searching the search key 55 of the word dictionary and examining the dictionary contents 56.

After the morphological analysis, the input text is sent to the pattern search unit 53, where the sentence pattern table 54 is searched and pattern matching is performed.

As described above, in this pattern table, the user of the system has previously registered sentences such as phrases specific to the specialized field as sentence patterns together with their bilingual sentences.

This sentence pattern will be explained in more detail below.

Figure 5 is a diagram showing the concept of sentence patterns, and Figure 5 (a
) is a diagram showing an actual sentence example, FIG. 5, etc.) is a diagram explaining the constituent elements of a sentence pattern, and FIG.

In other words, in the actual sentence example in Figure 5 (a), ``A long time ago, there lived an old man and an old woman in a place.'' ``The right old man'' and ``old woman'' are variables as noun-equivalent words. It is symbolized and converted into symbols N1 and N2, and the words in the (live and live) part are treated as selectable words.
Other words are fixed words.

Therefore, the constituent elements of the sentence pattern, as shown in Figure 5(b), consist of a fixed part, a variable part, and a selection part. The part〉 is a fixed fixed sentence, the variable part is an undetermined part encoded as a variable, and is ignored during pattern matching and determined at a later stage using a word dictionary. The variable selection section is a section that generates a plurality of modified sentence patterns by selecting one of a plurality of words.

In addition, in an actual sentence example, as shown in Figure 5(e),
The order of the fixed part, variable part, and variable selection part can be changed in various ways.

A pattern candidate list is extracted by searching a pattern table in which sentence patterns are registered.

[Problems to be Solved by the Invention] As explained above, in the conventional method, the search in the sentence pattern table and the pattern matching are performed separately from the morphological analysis. , it was necessary to perform the same test again in pattern search, which was inefficient.

The present invention has been made in view of the above-mentioned problems, and an object of the present invention is to provide a machine translation system that can simultaneously perform pattern search during morphological analysis and improve processing efficiency.

[Means for Solving the Problems] According to the present invention, the above objects are achieved by the means described in the claims.

That is, the present invention is a machine translation system that converts an input sentence in a source language into an output sentence in a target language. In a machine translation system that performs pattern matching between a sentence pattern registered in advance in the system and an input sentence in the source language, means for generating a sentence pattern dictionary using predetermined words in the sentence pattern as a search key. and a means for searching the pattern dictionary at the same time as searching the word dictionary during morphological analysis when analyzing an input sentence in the source language.

[Function] In the machine translation processing system of the present invention, during morphological analysis, the dictionary is looked up while decomposing the notation in the human sentence.
By setting the search key of the pattern dictionary as part of the notation to be compared in the same way as the word dictionary, the pattern dictionary can be searched simultaneously with the word dictionary using morphological analysis.

This will be explained below using a specific example.

Match the input sentence with the sentence pattern by matching the fixed part in the sentence pattern, matching with any of the selected variable part, and matching the remaining notation with the variable symbol part! It is established by acknowledging it. Therefore, for the input sentence to be pattern matched, the fixed pattern part must always match, so the first
As shown in the diagram illustrating the concept of the pattern dictionary structure, an example of the sentence pattern in Figure (a) is ``In a place long ago, N
1 and N2 (lived together)'', as shown in the structure of the pattern dictionary in figure (b), use the search key to search for part of the fixed part (``in a place long ago''). 1, and the other parts as pattern content 2 (r#N1 and N2 (lived) together.'') to create the same structure as the word dictionary shown in Figure 4.

This makes it possible to search for sentence patterns that may be matched during morphological analysis.

[Embodiment] FIG. 2 is a diagram illustrating an embodiment of the present invention, and shows only the source language input sentence analysis unit that is directly related to the present invention. In the figure, 3 is a morphological analysis section, 4 is a word dictionary, 5 is a pattern dictionary, and 6 is a section for sifting the word list and pattern candidate list.

That is, in this embodiment, the pattern dictionary 5 explained in FIG. , the word dictionary 4 and the pattern dictionary 3 are searched and collated at the same time, and a word list or a pattern candidate list can be obtained through the sieve section 6.

[Effects of the Invention] As explained above, according to the machine translation system of the present invention, search and matching of sentence patterns are performed simultaneously during morphological analysis, so that processing efficiency is improved accordingly.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the concept of a pattern dictionary structure, FIG. 2 is a diagram explaining an embodiment of the present invention, FIG. 3 is a diagram explaining a conventional translation processing method using a sentence pattern table, and FIG. FIG. 4 is a diagram showing the concept of word dictionary structure, and FIG. 5 is a diagram showing the concept of sentence pattern. 1... Search key in the sentence pattern dictionary 2...
...Pattern contents in the sentence pattern dictionary, 3...
・Morphological analysis section, 4... Word dictionary, 5...
...Pattern Dictionary, 6...The sifting of the word list and pattern candidate list is

Claims (1)

[Scope of Claims] A machine translation system that converts an input sentence in a source language into an output sentence in a target language, which performs morphological analysis using a word dictionary of the source language when parsing the input sentence in the source language. In addition to doing
In a machine translation system that performs pattern matching between sentence patterns registered in the system in advance and input sentences in the source language,
means for generating a sentence pattern dictionary using a predetermined word in the sentence pattern as a search key; What is claimed is: 1. A machine translation system having a sentence pattern dictionary, characterized in that: