JP3873653B2 - Dictionary creating method and apparatus, dictionary creating program, and storage medium storing dictionary creating program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a dictionary creation method and apparatus, a dictionary creation program, and a storage medium storing the dictionary creation program, and in particular, adds meanings described in a national language dictionary or the like to a machine translation dictionary (Japanese-English dictionary or the like). Creating a dictionary for building different language pair dictionaries (such as the Japanese-Japanese dictionary) by combining this with other translation dictionaries (such as the English horse (Malay) dictionary). The present invention relates to a method and apparatus, a dictionary creation program, and a storage medium storing the dictionary creation program.
[0002]
[Prior art]
Conventionally, as a method of mechanically creating a dictionary, a method of creating a dictionary using a bilingual corpus obtained by digitizing the same document written in different languages has been mainly used. In particular, a bilingual corpus translated directly between two languages such as Japanese and English can extract bilinguals with high accuracy.
[0003]
[Problems to be solved by the invention]
However, there are few types and amounts of bilingual corpora, and it is difficult to obtain a sufficient amount for creating a large-scale dictionary that can withstand the bilingual corpora of various language pairs.
[0004]
On the other hand, with the explosive spread of the Internet, machine translation and information retrieval techniques (including retrieval in different languages) are becoming increasingly important. The importance of “dictionaries” that support these technologies at the bottom is also increasing. It is common to create a dictionary manually, but there is a big problem that it takes cost and time to create a large-scale dictionary.
[0005]
The present invention has been made in view of the above points, and a dictionary creation method and apparatus and dictionary creation program capable of efficiently and highly accurately expanding dictionary information using a plurality of different electronic dictionaries. And a storage medium storing a dictionary creation program.
[0006]
[Means for Solving the Problems]
FIG. 1 is a diagram for explaining the principle of the present invention.
[0007]
The present invention (claim 1) is a dictionary creation process in the dictionary creating apparatus for creating a build electronic dictionaries and other different-language from the electronic dictionary electronic dictionary,
A first dictionary creation means comprising a language-to-dictionary comprising a first natural language headword, a semantic category, and a second natural language bilingual expression (translation), and a part of speech specified; Using a semantic dictionary composed of headwords and meanings, matching is performed based on the headwords of both dictionaries, and based on the headwords that have been matched, semantic categories and translated words are extracted from the language pair dictionary. To obtain at least a headword, reading, part of speech, meaning, and create a first dictionary and store it in the first storage means (step 1),
The morpheme analyzing means performs morphological analysis on the meaning of the first dictionary of the first storage means, and assigns a semantic category to the constituent words decomposed into morphemes (step 2),
The second dictionary creating means extracts the constituent words to which the semantic category is given by morphological analysis, combines with the first dictionary, creates a second dictionary, and stores it in the second storage means (step 3),
The similarity calculation means calculates the similarity by comparing the semantic category of the headword in the second dictionary of the second storage means with the semantic category of the meaning word obtained by morphological analysis, and calculating the similarity And the second dictionary are combined to create a third dictionary, which is stored in the third storage means (step 4),
Based on the similarity , the valid information extraction unit sets a threshold for the similarity of the third dictionary of the third storage unit, extracts a record having a similarity greater than or equal to the threshold (step 5), and is extracted The effective information is extracted by specifying the part of speech of the record, and the extracted effective information is output as a target dictionary (step 6).
[0008]
In the present invention (Claim 2), when calculating the similarity,
The similarity calculating means calculates all the similarities of the semantic category of the entry word of the language pair dictionary and the semantic category of the word obtained from at least the meaning of the entry word of the semantic dictionary, and calculates the maximum value of the similarity Similarity.
[0009]
FIG. 2 is a principle configuration diagram of the present invention.
[0010]
The present invention (Claim 3) is a dictionary creation device for creating an electronic dictionary of different language pairs from a constructed electronic dictionary and another dictionary,
A language-to-dictionary 200 composed of a headword of the first natural language, a semantic category, and a bilingual expression (translated word) of the second natural language, and a part of speech is specified;
A meaning dictionary 300 composed of headwords and meanings of the first natural language;
Matching is performed based on the headwords in the language vs dictionary and the meaning dictionary, and for the matched headword, the semantic category and translation are obtained from the language vs dictionary, and at least the headword, reading, part of speech, and meaning are obtained from the meaning dictionary . First dictionary creating means 110 for creating one dictionary 120;
A morpheme analysis unit 10 that performs morphological analysis on the meaning of the first dictionary 120 and assigns a semantic category to the constituent words decomposed into morphemes;
A second dictionary creating means 130 for extracting a constituent word given a semantic category by morphological analysis and combining it with the first dictionary to create a second dictionary 140;
Comparing the semantic category of the headword in the second dictionary 140 with the semantic category of the semantic word obtained by morphological analysis, calculating the similarity, and combining the similarity with the second dictionary 140; Similarity calculation means 150 for creating a third dictionary 160;
Based on the similarity, a threshold is set for the similarity of the third dictionary 160, a record having a similarity greater than or equal to the threshold is extracted, and effective information is extracted by specifying the part of speech of the extracted record, a valid information extracting means 170 for outputting the extracted the active information as a dictionary for the purpose, the.
[0011]
According to the present invention (Claim 4), in the similarity calculation means 150,
Means for calculating all similarities between the semantic category of the entry word of the language versus dictionary and at least the semantic category of the word obtained from the meaning of the entry word of the meaning dictionary, and setting the maximum value of the similarity as similarity .
[0012]
The present invention (Claim 5) is a dictionary creation program for causing a computer to execute each means of the dictionary creation apparatus according to Claim 3 or 4.
[0014]
The present invention (Claim 6 ) is a storage medium storing a program for causing a computer to execute each means of the dictionary creation device according to Claim 3 or 4.
[0016]
There is a dictionary (“Japanese Vocabulary System” Iwanami Shoten) that categorizes Japanese words into approximately 3000 semantic categories as an accumulation of linguistic knowledge obtained in the process of machine translation research and development. There is also a large Japanese-English dictionary created based on this semantic category. In the present invention, such semantic categories of words are applied to the creation of a dictionary. The present invention includes a “language pair dictionary” composed of “headwords”, “semantic categories” of a first natural language, and bilingual expressions (“parallel translations”) of a second natural language, and having a part of speech specified. Based on each headword using a "meaning dictionary" composed of natural headlines, meanings (ready, part of speech, example sentences, antonyms, synonyms, synonyms, etc.) To create a new “intermediate dictionary”. A morphological analysis is performed on the created meaning of the “intermediate dictionary” (part corresponding to the meaning of the meaning dictionary) and a semantic category is assigned to the constituent words. Next, the semantic category of “intermediate dictionary” (the portion corresponding to the semantic category of the language versus dictionary) is compared with the semantic categories of the constituent words obtained by morphological analysis, and the similarity is calculated. A target “dictionary” can be created by extracting records having a degree of similarity equal to or greater than a certain threshold from the “intermediate dictionary” based on the obtained similarity and further specifying a part of speech.
[0017]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below.
[0018]
FIG. 3 shows the configuration of the dictionary creation apparatus of the present invention.
[0019]
The dictionary creation device 100 shown in the figure includes a language pair dictionary 200, a semantic dictionary 300, a morphological analyzer 10, a first auxiliary storage device 120, a second auxiliary storage device 140, a third auxiliary storage device 160, and A dictionary storage device 400 for storing the created dictionary is connected.
[0020]
The dictionary creation device 100 includes a dictionary entry combining unit (intermediate dictionary 1 creation unit) 110, an intermediate dictionary 2 creation unit 130, a similarity calculation unit (intermediate dictionary 3 creation unit) 150, and an effective information extraction unit 170.
[0021]
The language pair dictionary 200 includes a “language pair dictionary” that is composed of “headwords” in the first natural language, “semantic categories”, and bilingual expressions in the second natural language (“translated words”), and in which the part of speech is specified. Is stored. It is assumed that the language pair dictionary 200 is stored in a storage device such as a hard disk.
[0022]
The semantic dictionary 300 is composed of “headwords” and “meanings” (part of speech, example sentences, antonyms, synonyms, synonyms, etc.) in the first natural language. It is assumed that the semantic dictionary 300 is stored in a storage device such as a hard disk.
[0023]
The dictionary entry coupling unit 110 of the dictionary creation device 100 reads dictionary information from the language pair dictionary 200 and the semantic dictionary 300, performs matching by the respective entry words, creates the intermediate dictionary 1, and stores it in the first auxiliary storage device 120. Store.
[0024]
For the intermediate dictionary 1 stored in the first auxiliary storage device 120, the morpheme analyzer 10 with the semantic category assignment function uses the meaning part of the intermediate dictionary 1 (the part corresponding to the meaning of the meaning dictionary 300). The morpheme analysis is performed, and the morpheme analysis result is passed to the intermediate dictionary creation unit 130.
[0025]
The intermediate dictionary 2 creation unit 130 extracts the word salmon to which the semantic category is assigned based on the information morphologically analyzed by the morphological analyzer 10, and the intermediate dictionary 1 stored in the first auxiliary storage device 120. The intermediate dictionary 2 is created by combining them and stored in the second auxiliary storage device 140.
[0026]
The similarity calculation unit 150 is obtained from the semantic category of the intermediate dictionary 2 stored in the second auxiliary storage device 140 (part corresponding to the semantic category of the language pair dictionary 200) and the morphological analyzer 10 with a semantic category assignment function. The degree of similarity of the semantic categories of the word meaning constituent words is calculated, and the result is combined with the intermediate dictionary 2 to create the intermediate dictionary 3 and stored in the third auxiliary storage device 160.
[0027]
The valid information extraction unit 170 extracts valid information from the intermediate dictionary 3 stored in the third auxiliary storage device 160 based on the similarity and the part of speech, creates a dictionary, and stores the dictionary in the dictionary storage device 400.
[0028]
The intermediate dictionary 1 stored in the first auxiliary storage device 120 includes headwords, semantic categories, translations, readings, parts of speech, meanings, and example sentences.
[0029]
The intermediate dictionary 2 stored in the second auxiliary storage device 140 includes headwords, semantic categories, translations, readings, parts of speech, meanings, example sentences, and morphological analysis results.
[0030]
The intermediate dictionary 3 stored in the third auxiliary storage device 160 includes headwords, semantic categories, translated words, readings, parts of speech, meanings, example sentences, morphological analysis results, and similarities. The target dictionary stored in the dictionary storage device 400 includes headwords, semantic categories, translations, readings, parts of speech, meanings, example sentences, morphological analysis results, and similarities.
[0031]
It is assumed that the morphological analyzer 10 uses an existing one with a semantic category assignment function.
[0032]
Next, the operation of the dictionary creating apparatus having the above configuration will be described.
[0033]
FIG. 4 is a flowchart of dictionary creation processing according to an embodiment of the present invention.
[0034]
Step 101) The dictionary entry combining unit 110 performs matching based on the respective headwords of the language pair dictionary 200 and the semantic dictionary 300, and based on the headwords that can be matched, for example, from the language pair dictionary 200, meaning categories and translations Are acquired from the semantic dictionary 300, the intermediate dictionary 1 is created by acquiring headwords, readings, parts of speech, meanings and example sentences, and stored in the first auxiliary storage device 120.
[0035]
Step 102) Take out the meaning part of the intermediate dictionary 1 stored in the first auxiliary storage device 120 (if there are “example sentence”, “opposite words”, “synonyms”, “synonyms”, etc.), Is input to the external morpheme analyzer 10 with a semantic category assignment function, and the constituent words decomposed into morphemes and data to which the “semantic category” is assigned are received.
[0036]
Step 103) Thereafter, only the words to which the “semantic category” is assigned are extracted from the constituent words decomposed into morphemes by the intermediate dictionary 2 creation unit 130, and the intermediate words stored in the first auxiliary storage device 120 are extracted. It is combined with the dictionary 1 and stored in the second auxiliary storage device 140 as the intermediate dictionary 2. For example, the contents of the intermediate dictionary 2 are composed of headwords, semantic categories, translations, readings, parts of speech, meanings, example sentences, and morphological analysis results.
[0037]
Step 104) The similarity calculation unit 150 calculates the “similarity” for each “meaning category” of the word included in the semantic category and the morphological analysis result of the intermediate dictionary 2 stored in the second auxiliary storage device 140, The obtained calculation result is combined with the intermediate dictionary 2 and stored in the third auxiliary storage device 160 as the intermediate dictionary 3. For example, the content of the intermediate dictionary 3 includes a headword, a semantic category, a translation, a reading, a part of speech, a meaning, an example sentence, a morphological analysis result, and a similarity.
[0038]
Step 105) The valid information extraction unit 170 sets a threshold value for the similarity of the intermediate dictionary 3 stored in the third auxiliary storage device 160, further extracts the valid information by narrowing down the part of speech and the like, and the dictionary storage device 400 is stored.
[0039]
【Example】
Embodiments of the present invention will be described below with reference to the drawings.
[0040]
In the following example, it is assumed that the headword is “dog”, the first natural language is “Japanese”, and the second natural language is “English”.
[0041]
FIG. 5 shows an example of a language pair dictionary according to an embodiment of the present invention.
[0042]
In the language pair dictionary 200 shown in the figure, the headword “dog” is registered in two ways: a semantic category “316 spy”, a translated word “spy”, a semantic category “537 beast”, and a translated word “dog”. Although not shown, the part of speech of the headword “dog” is “noun”.
[0043]
FIG. 6 shows an example of the meaning dictionary according to one embodiment of the present invention. As shown in the figure, in the meaning dictionary 300, the headword “dog” is composed of columns of “yomi”, “part of speech”, “meaning”, and “example sentence”, including those whose reading is “ken”. 5 types are registered.
[0044]
When the dictionary entry combination unit 110 performs matching using a headword using the language pair dictionary 200 and the semantic dictionary 300, an intermediate dictionary 1 as shown in FIG. 7 is created. The intermediate dictionary 1 is created with 10 records because there are two pairs of language pairs dictionary 200 and five ways of meaning dictionary 300 registered. The created intermediate dictionary 1 is stored in the first auxiliary storage device 120.
[0045]
Next, the “meaning” and “example sentence” portions are extracted from the intermediate dictionary 1 (if there are “opposite words”, “synonyms”, “synonyms”, etc.), and they are converted into morphological analysis with an external semantic category assignment function. Input to the device 10. The morphological analysis device 10 has a function of assigning, for example, a semantic category of a word (noun) classified by “Japanese vocabulary system” (Iwanami Shoten) to a constituent word of an input sentence. Using the result obtained from the morpheme analyzer 10, if only the words to which the “semantic category” is assigned are extracted, the information shown in the right column of FIG. 8 is obtained. When this is combined with the aforementioned intermediate dictionary 1 (FIG. 7), an intermediate dictionary 2 as shown in the upper column of FIG. 9 is obtained.
[0046]
Using this intermediate dictionary 2, the similarity calculation unit 150 uses the “meaning category” included in the entry word of the language pair dictionary 200 and the “meaning” and “example sentence” included in the entry word of the meaning dictionary 300. The similarity of the “semantic category” of the constituent words obtained from is calculated. The result is shown in the lower column of FIG. A method for calculating the similarity will be described later.
[0047]
Next, when the similarity calculation result obtained by the similarity calculation unit 150 is combined with the intermediate dictionary 2 in the upper column of FIG. 9, the intermediate dictionary 3 shown in FIG. 10 is created.
[0048]
Finally, the valid information extraction unit 170 sets a threshold value for the similarity of the intermediate dictionary 3 (0.8 or more in this example), further narrows down the part of speech to “noun” (the entry word of the language versus dictionary 200). If the valid information is extracted, the dictionary as shown in FIG. 11 is created and stored in the dictionary storage device 400 (because the part of speech is “noun”).
[0049]
Note that the record whose reading is “Ken” and the part of speech is “suffix” is deleted in this example because the part of speech is different, although the similarity is larger than the threshold.
[0050]
Next, similarity calculation in the similarity calculation unit 150 will be described.
[0051]
FIG. 16 is a flowchart of similarity calculation according to an embodiment of the present invention.
[0052]
First, the semantic category number of the input headword is acquired (step 201), and the number of constituent words of the input meaning is acquired (step 202).
[0053]
If there is no constituent word (step 203, no), the maximum value of the similarity is obtained for each constituent word, and the process ends (step 207).
[0054]
If there is still a constituent word (step 203, yes), the semantic category number of the constituent word is acquired (step 204).
[0055]
If there is still a semantic category (step 205, yes), the similarity (distance) between the semantic category number of the headword and the semantic category number of the constituent word is calculated (step 206), and the process proceeds to step 205, where the semantic category If there is no (step 205, no), the process proceeds to step 203.
[0056]
In the following example, the headword “dog”, semantic category “316 spy”, meaning (morphological analysis result) “turner ([316, 347]), spy ([316, 347, 1427]) are input as input. Shall be.
[0057]
The semantic categories of words (nouns) recorded in “Japanese Vocabulary System” (Iwanami Shoten) have a tree structure as shown in FIGS. In FIG. 12, the tree structure of “spy” is shown as the semantic category, and in FIG. 13, the tree structure of “beast” is shown as the semantic category.
[0058]
When attention is paid to the semantic categories “316 spy” and “537 beast” of the headword “dog” in the language versus dictionary 200, the semantic category includes words as shown in FIGS.
[0059]
When this constituent word is examined, the words “Inu”, “Spy”, and “Rotator” obtained from the “meaning” and “example sentence” that the headword of the meaning dictionary 300 has have the meaning category “316 spy”. It can be seen that “Inu”, “Japanese dog”, and “Guide dog” are registered as constituent words of the semantic category “537 beast”. In this way, when there is a phrase that matches the meaning category of the entry word in the language pair dictionary 200 and the meaning category of the word obtained from the “meaning” and “example sentence” that the entry word of the meaning dictionary 300 has, The similarity is given as “1.0000000”. The number of appearances is also given.
[0060]
On the other hand, when there is no match at all, for example, the meaning of the meaning word “316” of the headword “dog” in the language vs. dictionary 200 is “don't care carnivore. Although the degree of similarity is “0.400000”, when attention is paid to the constituent word “owner ([349])” of this meaning, as can be seen from FIG. 12, the semantic category “316 spy” and the semantic category “349 possession”. It can be understood that “person” (owner's semantic category) is 316 → 311 → 224 → 223 → 333 → 348 → 349 and distance “6” when tracing the tree.
[0061]
this,
1−0.1 * N N = 0 to 9: Distance of semantic category between words 1 / (N + 1) N ≧ 10: Calculated by applying to distance of semantic category between words, and set to “0.400000”.
[0062]
In any case, all the similarities between the semantic categories of the headwords in the language pair dictionary 200 and the semantic categories of the words obtained from the “meaning” and “example sentences” of the headwords in the semantic dictionary 300 are calculated, The maximum value is adopted.
[0063]
In addition, when there is a word whose notation matches the “entry word” among the words obtained from the “meaning” and “example sentence” that the entry word of the meaning dictionary 300 has, the obtained similarity is adversely affected. Therefore, the word is not subject to similarity calculation (for example, “Dog” obtained from “Dog Doctor” in FIG. 9).
[0064]
In the above example, the components of the dictionary creation device 100 shown in FIG. 3 can be constructed as a program and installed in the CPU of a computer used as the dictionary creation device, or distributed via a network. is there.
[0065]
In addition, the constructed program is stored in a portable storage medium such as a hard disk, a floppy disk, or a CD-ROM connected to a computer used as a dictionary creation device, and is installed when the present invention is carried out. Thus, the present invention can be easily realized.
[0066]
The present invention is not limited to the above-described embodiments, and various modifications and applications are possible within the scope of the claims.
[0067]
【The invention's effect】
As described above, according to the present invention, it is possible to efficiently expand an electronic dictionary using a plurality of existing electronic dictionaries. For example, in combination with target dictionaries that have meanings in Japanese-English dictionaries (English, French, English, Chinese, etc.), use the similarities of meanings of entries in each target dictionary, and target dictionaries for new language pairs (Japanese-French , Daytime, day horse, etc.) can be created automatically.
[Brief description of the drawings]
FIG. 1 is a diagram for explaining the principle of the present invention.
FIG. 2 is a principle configuration diagram of the present invention.
FIG. 3 is a configuration diagram of a dictionary creation device according to an embodiment of the present invention.
FIG. 4 is a flowchart of dictionary creation processing according to an embodiment of the present invention.
FIG. 5 is an example of a language pair dictionary according to an embodiment of the present invention.
FIG. 6 is an example of a meaning dictionary according to an embodiment of the present invention.
FIG. 7 is an example of the created intermediate dictionary 1 according to an embodiment of the present invention.
FIG. 8 shows information to be input to the morphological analyzer according to one embodiment of the present invention and the obtained result.
FIG. 9 is a diagram showing the similarity with the intermediate dictionary 2 of one embodiment of the present invention.
FIG. 10 is an example of the created intermediate dictionary 3 according to an embodiment of the present invention.
FIG. 11 is an example of a dictionary created in one embodiment of the present invention.
FIG. 12 shows a tree structure (Japanese vocabulary system) of a semantic category “spy” according to an embodiment of the present invention.
FIG. 13 shows a tree structure (Japanese vocabulary system) of the semantic category “beast” according to one embodiment of the present invention.
FIG. 14 is a constituent word (Japanese vocabulary system) of a semantic category “spy” according to an embodiment of the present invention.
FIG. 15 is a constituent word (Japanese vocabulary system) of a semantic category “beast” according to an embodiment of the present invention.
FIG. 16 is a flowchart of similarity calculation according to an embodiment of the present invention.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 10 Morphological analyzer 100 Dictionary creation apparatus 110 1st dictionary creation means, dictionary entry combination part 120 1st dictionary, 1st auxiliary storage device 130 2nd dictionary creation means, intermediate dictionary 2 creation part 140 2nd dictionary , Second auxiliary storage device 150 similarity calculation unit, similarity calculation means 160 third dictionary, third auxiliary storage device 170 effective information extraction means, effective information extraction unit 200 language pair dictionary 300 semantic dictionary 400 Dictionary, dictionary storage

Claims (6)

A dictionary creation process in the dictionary creating apparatus for creating a build electronic dictionaries and other different-language from the electronic dictionary electronic dictionary,
A first dictionary creation means comprising a language-to-dictionary comprising a first natural language headword, a semantic category, and a second natural language bilingual expression (translation), and a part of speech specified; Using a semantic dictionary composed of headwords and meanings, matching is performed based on the headwords of both dictionaries, and based on the headwords that have been matched, semantic categories and translated words are extracted from the language pair dictionary. To obtain at least a headword, a reading, a part of speech, a meaning, and create a first dictionary, which is stored in a first storage means,
The morpheme analyzing means performs morphological analysis on the meaning of the first dictionary of the first storage means, and assigns a semantic category to the constituent words decomposed into morphemes,
The second dictionary creating means extracts a constituent word to which a semantic category is given by morphological analysis, combines it with the first dictionary, creates a second dictionary, and stores it in the second storage means;
The similarity calculation means calculates the similarity by comparing the semantic category of the headword in the second dictionary of the second storage means with the semantic category of the meaning word obtained by the morphological analysis, Combining the similarity and the second dictionary to create a third dictionary and storing it in a third storage means;
Based on the similarity , the valid information extraction unit sets a threshold for the similarity in the third dictionary of the third storage unit , extracts a record having a similarity greater than the threshold, and is extracted A dictionary creation method, wherein valid information is extracted by specifying a part of speech of the record, and the extracted valid information is output as a target dictionary. When calculating the similarity,
The similarity calculation means calculates all similarities between the semantic category of the entry word of the language pair dictionary and the semantic category of the word obtained from at least the meaning of the entry word of the semantic dictionary, and the maximum of the similarity degree The dictionary creation method according to claim 1, wherein the value is a similarity. A dictionary creation device for creating an electronic dictionary of a different language pair from a constructed electronic dictionary and another dictionary,
A language-to-dictionary comprising a first natural language headword, a semantic category, and a second natural language bilingual expression (translation), and a part of speech specified;
A lexicon of words composed of headwords and meanings of the first natural language;
Matching is performed based on the headwords of the language pair dictionary and the semantic dictionary, and for the headwords that can be matched, the semantic category and the translated word from the language pair dictionary, and at least the headword, reading, part of speech, and meaning from the semantic dictionary a first dictionary creation means for creating acquired by the first dictionary,
A morpheme analysis unit that performs morphological analysis on the meaning of the first dictionary and assigns a semantic category to the constituent words decomposed into morphemes;
Second dictionary creating means for extracting a constituent word to which a semantic category is given by morphological analysis and combining it with the first dictionary to create a second dictionary;
Comparing a semantic category of a headword in the second dictionary with a semantic category of a semantic word obtained by the morphological analysis, calculating a similarity, and combining the similarity with the second dictionary; Similarity calculating means for creating a third dictionary;
Based on the similarity, a threshold is set for the similarity in the third dictionary, a record having a similarity greater than or equal to the threshold is extracted, and valid information is extracted by specifying the part of speech of the extracted record Effective information extraction means for outputting the extracted effective information as a target dictionary;
A dictionary creating apparatus characterized by comprising: The similarity calculation means includes:
Means for calculating all the similarities between the semantic category of the entry word of the language pair dictionary and the semantic category of the word obtained from at least the meaning of the entry word of the semantic dictionary, and setting the maximum value of the similarity as the similarity The dictionary creation device according to claim 3. On the computer,
5. A dictionary creation program that causes each means of the dictionary creation device according to claim 3 to be executed . On the computer,
5. A storage medium storing a dictionary creating program, wherein a program for executing each means of the dictionary creating apparatus according to claim 3 is stored.

Images