JPH05233686A - Japanese language processor - Google Patents

Abstract

PURPOSE:To provide a Japanese language processor which reduces remarkably a load of user's dictionary registration, and is excellent particularly in appropriateness of an analysis and appropriateness of a translated word in the case it is used for a translation. CONSTITUTION:The processor is constituted of a mechanical processing dictionary 1 for eliminating registration of a proper noun, a morpheme analyzing means 2 for dividing and writing an input sentence into morphemes by using this dictionary 1, a morpheme analytic result check table 3 for defining a certain specific morpheme analytic result as unnatural such as it becomes an undefined word section, or is divided and written in one-character KANJI (Chinese character) continuation, a proper noun sequence word relation qualification table 4 for holding information of an affix appearing before and behind a proper noun, a dictionary re-retrieval candidate section qualifying means 5 for qualifying a section for consulting a proper noun dictionary 6 by utilizing these two tables, the proper noun dictionary 6, and a proper noun dictionary retrieving means 7.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a Japanese language processor,
In particular, a large number of dictionaries that need to be registered in the dictionary adversely affect the morpheme splitting process of the input sentence, so that it is difficult to handle the Japanese noun dictionary, which is difficult to handle. Regarding

[0002]

2. Description of the Related Art In a conventional Japanese language processing apparatus, registration of proper nouns in a dictionary is left to the user like other unknown words. The reason why proper nouns are not treated in the system dictionary is that the number of words is large, the hit rate per word is low, that is, the ratio of the number of different occurrence words to the total number of appearance words is close to 1, and above all It is said that existence has a bad influence on the separation of input sentences.

However, even in a manual sentence that is said to be suitable for machine processing, the appearance frequency of proper nouns cannot be ignored. Therefore, in order to reduce the burden on the user of registering proper nouns in the dictionary, several methods have been attempted for the conventional Japanese language processing apparatus.

One is a method of processing unregistered sections as proper nouns or numbers (Japanese Patent Laid-Open No. 62-280966). However, this method has a problem that the semantic attribute information of the part treated as a noun is uncertain.

As a method of solving this problem, it has been considered to collect attribute information of proper noun parts from information of words that are likely to appear before and after the proper noun (Japanese Patent Laid-Open No. 63-94365).
However, this method does not work if there is no synonym before and after.

[0006]

The conventional Japanese language processing apparatus described above has a problem that the presence of the proper noun heading adversely affects the result of sentence separation and a problem that the burden of registering the proper noun dictionary on the user increases. However, in general, the former is prioritized, and the latter problem is covered by means of estimating proper noun dictionary sections and their contents. But with this method,
Since there is no guarantee of dictionary information, for example, in the case of translation, there is no guarantee of translated words, and in the end, there is a problem that the necessary dictionary registration cost of proper nouns occurs on the user side.

It is an object of the present invention to first identify proper noun candidate sections and then look up proper nouns in the dictionary, thereby significantly reducing the load on the user's dictionary registration and adequacy of analysis. It is to provide a Japanese language processing device that guarantees the above and guarantees the appropriateness of the translated word at the time of translation.

[0008]

The Japanese language processing apparatus according to the first aspect of the present invention includes a first dictionary which eliminates registration of proper nouns, and
A second dictionary in which only proper nouns such as a person's name and a place name are registered, a morpheme analysis means for dividing an input character string into morpheme units registered in the first dictionary, and a result output from the morpheme analysis means And a dictionary re-search candidate section recognizing unit that determines a re-search candidate section according to a predetermined rule, and the second dictionary is subtracted from the candidate section output by the dictionary re-search candidate section recognizing unit. Iku proper noun dictionary search means.

A Japanese language processing apparatus of a second invention is the Japanese language processing apparatus of the first invention, and a morphological analysis result check table storing a plurality of definitions of unnatural morphological analysis results, and a proper noun synonym heading. The dictionary re-search candidate section recognizing means includes the proper noun synonym relation authorization table that stores information, its part-of-speech information, and semantic attribute information, and the dictionary re-search candidate section authorization means refers to the morphological analysis result check table and the proper noun synonym relation authorization table. Then, the re-search candidate section is determined and configured.

[0010]

Embodiments of the present invention will now be described with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the Japanese language processing apparatus of the present invention.

As shown in FIG. 1, the Japanese language processing apparatus of the present embodiment is a dictionary 1 for machine processing in which proper noun registration is eliminated.
And a morphological analysis unit 2 that divides an input sentence into morphemes by using this dictionary 1, and a certain morphological analysis result that is unnatural, such as an undefined word section, or one-character Kanji continuous division. A morphological analysis result check table 3, a proper noun synonym relation recognition table 4 that holds information on affixes that appear before and after a proper noun, and these two tables are used to identify a section in which the proper noun dictionary 6 is drawn. The dictionary re-search candidate section recognizing means 5, a proper noun dictionary 6, and a proper noun dictionary searching means 7 are included.

The morpheme analysis means 2 uses the dictionary 1 to divide a Japanese input sentence into morphemes which are the headings of the dictionary 1, and allocates dictionary information to each morpheme. It is considered that the dictionary re-search candidate section recognizing unit 5 should inquire the result of morpheme division and the dictionary information with the morpheme analysis result check table 3 and the proper noun synonymous word relationship recognition table 4 and try the dictionary lookup again. To certify.

The morphological analysis result check table 3 is a list of unnatural morphological analysis result definitions. The definitions are used for character types, morpheme lengths and their combinations, and whether or not there is a corresponding dictionary heading, for example: ・ Unregistered word section (section without corresponding dictionary heading) ・ For morphemes with two or more katakana strings Divided section ・ A section in which a 2-character Kanji string is divided into 1-character kanji morphemes ・ A section in which a 4-character Kanji string is divided into 1 character, 2 characters, and 1 character morphemes Defined or described.

The proper noun synonymous word relation recognition table 4 stores a set of three sets of headword information, part-of-speech information, and possible semantic attribute information of front / back suffix. FIG. 2 exemplifies the contents of the proper noun synonym relation recognition table 4.

The proper noun dictionary searching means 7 looks up the proper noun dictionary 6 for the section output by the dictionary re-search candidate section recognizing means 5, and if the search is successful, the result is a new morphological analysis and dictionary lookup result. If it fails, the output of the morphological analysis unit 2 is respected for that section.

The operation of this embodiment will be described with reference to FIG. When the sentence to be analyzed "See Yamada (Professor of Oxford University) for the following" is input (step 31), the morphological analysis means 2 makes the dictionary 1
The input sentence is divided into pieces by using (step 32). In the box descending from each morpheme of the segmentation result of step 32, the dictionary information of the dictionary lookup result of the dictionary 1 is stored (the dictionary information shown in the figure has extracted only necessary parts). ..

Since the proper noun heading is excluded from the dictionary 1, "Yamada" is divided into two morphemes, "Mountain" and "Ta," and "Oxford" is an unregistered word. The default dictionary information of unregistered words is stored in the box connected to. In addition, in a box connected from the morpheme of "university", information such as the possibility of nouns and the possibility of suffixes is stored.

With respect to the output of the morpheme analyzing means 2, the dictionary re-search candidate section recognizing means 5 first searches the morpheme analysis result check table 3 for a corresponding character string. The divided sentence obtained in step 32 includes the two-character Chinese character string “Yamada” and the unregistered word section “Oxford” which are divided into one-character morphemes.

The dictionary re-search candidate section recognizing means 5 uses those 2
After checking whether the words before and after one section are words that can be connected to the noun, these two character strings are held as re-search candidate sections. At that time, it is stored as a pair with information that the condition of the semantic attribute for the re-search candidate section is unlimited.

Next, the dictionary re-search candidate section recognizing means 5 checks whether or not the word corresponding to the heading of the proper noun synonymous word relation recognizing table 4 is included in the sentence. "University"
The two headings of “Professor” and “Professor” match the headings of the proper noun synonym relation recognition table 4, and the part-of-speech information held in “University” and “Professor” in step 32 corresponds to the proper noun syntactic relation recognition table 4. Meet the part of speech of the entry.

The dictionary re-search candidate section recognizing means 5 extracts a character string in which the character type is changed from the character immediately before each word to the left. As a result, "Oxford" and "University" are recognized as candidate words for the suffixes "University" and "Professor", respectively.

The dictionary re-search candidate section recognizing means 5 holds the candidate character strings and the semantic attribute information "place, organization" and "person" obtained from the proper noun synonym relation recognition table 4 in pairs. At this time, for the information pair related to the character string “Oxford”, the result already obtained by referring to the morphological analysis result check table 3 is overwritten.

This is because the section recognized by the proper noun synonym relation recognition table 4 is restricted in the semantic attribute, so that the proper noun dictionary 6 can be retrieved under more strict conditions. .. As a result, the output of the dictionary re-search candidate section recognizing means 5 is as follows.

(Character string) (Semantic attribute condition) Yamada ANY Oxford Place, organization University person The above "ANY" means "the condition of the semantic attribute is unlimited".

The proper noun dictionary searching means 7 looks up the proper noun dictionary 6 based on this output result, and a dictionary having a meaning attribute of "person" in "Yamada" and a dictionary having a meaning attribute of "place" in "Oxford". And search successfully. But "university"
The search fails because it does not exist in the proper noun dictionary 6 in the first place. The dictionary search means 7 checks that the semantic attribute information in the "Yamada" and "Oxford" dictionaries meets the conditions of the above output result, and then the morpheme division result and the dictionary information for the two character string sections are searched. Replacement is performed (step 33).

In this way, in this embodiment, the appropriateness of the Japanese analysis and the appropriateness of the translated word in the case of translation are guaranteed.

[0028]

As described above, the Japanese language processing apparatus of the present invention performs the two-step processing of recognizing proper noun candidate sections and pulling the proper noun dictionary, thereby adversely affecting registration of the proper noun dictionary. At the same time, it is possible to eliminate the risk of inferring proper noun information not found in the dictionary.

Therefore, there is an effect that the load of the user's dictionary registration is reduced, the appropriateness of analysis is guaranteed, and the appropriateness of the translated word is guaranteed when used for translation.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a Japanese language processing apparatus of the present invention.

FIG. 2 is a diagram illustrating a data configuration of a proper noun synonym relation authorization table.

FIG. 3 is a flowchart showing a processing operation of this embodiment.

[Explanation of symbols]

 1 dictionary 2 morpheme analysis means 3 morpheme analysis result check table 4 proper noun synonym relation recognition table 5 dictionary re-search candidate section recognition means 6 proper noun dictionary 7 proper noun dictionary search means 31, 32, 33 flow chart steps

Claims (2)

[Claims] 1. A first dictionary excluding registration of proper nouns, a second dictionary in which only proper nouns such as person names and place names are registered, and an input character string is registered in the first dictionary. A morpheme analysis unit that divides into morpheme units, a dictionary re-search candidate section determination unit that determines a re-search candidate section according to a predetermined rule for the result output from the morpheme analysis unit, and the dictionary re-search candidate A Japanese language processing apparatus comprising: a proper noun dictionary searching unit that refers to the second dictionary for a candidate section output by the section recognizing unit. 2. A morphological analysis result check table that stores a plurality of unnatural morphological analysis result definitions, and a proper noun synonym relation recognition table that stores heading information of proper noun synonyms, part of speech information, and semantic attribute information. 3. The dictionary re-search candidate section recognizing means refers to the morpheme analysis result check table and the proper noun synonymous word relationship recognition table, and the dictionary re-search candidate section recognizing unit determines the re-search candidate section. Word processor.