JPH06208580A - Dictionary preparing device - Google Patents

Abstract

PURPOSE:To efficiently obtain data for a customized dictionary without allowing a user to be conscious of information in general use dictionary. CONSTITUTION:A sample text extracting part 1 extracts a sample text from a corpus 12 of a field to be processed. The extracted text is sent to an analysis information instructing part 2. The part 2 instructs analysis information to the sample text provided by a user. An analysis result is properly processed by an analysis result storage part 3 and stored in an analysis result file 13 or directly sent to the succeeding processing. A dictionary information specifying part 4 specifies information matched with the sample analysis result. The specified information is collected in each original unit by a dictionary information storage part 5. The stored information is synthesized as dictionary description by a dictionary synthesizing part 6 in accordance with a format stored in a format file 10 and stored in a customized dictionary file 11.

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to a dictionary creating apparatus, and more specifically, in natural language processing, information necessary for dictionary synthesis is extracted from a sample text analysis result, and a dictionary suitable for a processing target document is synthesized. The present invention relates to a dictionary creating device.

[0002]

2. Description of the Related Art In natural language processing such as machine translation,
Dictionary data that describes information about words in the processing target language is essential. In general-purpose dictionaries, in order to deal with a wide variety of linguistic phenomena, it is usual to describe information about polysemy of words and their proper use as detailedly and exhaustively as possible. However, when the fields or documents to be processed are limited, the polysemy is considerably eliminated, so that not only a large amount of properly used information is wasted, but also incorrect processing is given due to improper information addition. There are also many.

As a means for solving this problem, a method has been adopted in which the user directly registers the information on the word of the document to be processed by user dictionary registration or the like. For example, “Translation processing method and device with dictionary creation support function” proposed in Japanese Patent Laid-Open No. 2-67679 proposes information such as translated words by a user based on information such as the appearance frequency of words extracted from a sample text. Is selected from the general-purpose dictionaries, and the user dictionary is created based on that, that is, the user selects the necessary information from the general-purpose dictionary based on the frequency information in the target document. We are trying to reduce the burden on users and avoid registration mistakes. However, in the above-mentioned conventional technology, which of the polysemous words is to be selected is left only to the subjectivity of the user, and only the frequency information allows the user to objectively determine which word is used in the target document. The problem arises that there is no means to indicate.

[0004]

[Object] The present invention has been made in view of the above situation, and extracts a sample text of an appropriate quality and quantity from a document to be processed, and linguistically analyzes the sample text. To efficiently obtain customized dictionary data without paying attention to the information of the above, and especially as an analysis method, the sample sentence is divided into word units, and the modification target word and its relation name are specified for each word. By using the analysis and matching the information in the general-purpose dictionary according to the consistency of the dependency, the usage in the sample sentence can be automatically specified. Furthermore, if it is not possible to specify only the consistency of the dependency, the syntactic rule , Refinement of processing using means such as term structure matching and priority, and the information obtained from the sample sentence is the solid form unit appearing in the sentence, Up together into a distribution, it is a registrable information directly as dictionary data, also 1
When there is a possibility that there are a plurality of original forms for one solid form, it is possible to obtain accurate information by narrowing down the original form to one according to the specified information and the correspondence with the translated word specified by the user. By synthesizing the information obtained in step 1 according to a pre-defined format, and adding information specified by the user that is not in the general-purpose dictionary, or supplementing information that is not directly obtained from the sample sentence from the general-purpose dictionary, The object of the present invention is to provide a dictionary creation device that creates a customized dictionary suitable for processing a target document.

[0005]

In order to achieve the above object, the present invention provides (1)
A sample text extraction unit that extracts a sample text from a document of a processing target field, an analysis information instruction unit that presents the extracted sample text and instructs analysis information of a sentence in the text, and stores the instructed analysis information. An analysis result storage unit, and a dictionary information specifying unit that specifies information suitable for usage in the sample for each word in the sample sentence based on the analysis result stored in the analysis result storage unit from information in a general-purpose dictionary, Based on the information specified in the dictionary information specification unit, the dictionary information storage unit that stores frequency information such as the frequency of appearance of words, and the prescribed description format, while referring to the information of the general-purpose dictionary, if necessary, A dictionary synthesizing unit for synthesizing the information stored in the dictionary information storage unit into a dictionary format, and (2) information to be instructed by the analysis information instructing unit. Is a relationship destination of each word in the sample sentence and its qualified relation name, and based on the information, in the analysis result storage unit, a qualified destination, a qualified source, and each qualified relationship name are stored for each word, and the dictionary is stored. The information specifying unit has a specifying unit for specifying information matching the dependency matching among the dictionary lookup information by the general-purpose dictionary for each word, and (3) the dictionary in (2) above. In the information specifying unit, if unsatisfactory information in the general dictionary cannot be eliminated only by the consistency of dependency,
Having an excluding means for excluding inappropriate information by applying syntactic rules, and (4) In (2) above,
The dictionary information specifying unit specifies the term structure such as a verb in the sample sentence according to the number and type of terms extracted from the dependency information, and extracts the information matching the specified term structure from the information of the general-purpose dictionary. (5) In (3) or (4), the dictionary information specifying unit cannot specify the information suitable for the usage in the sample sentence from the information of the general dictionary by the removing unit and the extracting unit. In this case, there is provided means for specifying information according to a priority order arbitrarily determined according to the nature of the language of the sample text, and (6) in (1) or (2) above, the dictionary information storage The section has means for collecting and listing the dictionary information obtained in units of the solid form in the sample sentence as information in units of the original form in the general-purpose dictionary, and further (7 In the above (1), when there is a possibility that there are a plurality of original forms for one solid word form in the dictionary information storage unit, selecting means for selecting an original form that matches the information specified by the dictionary information specifying unit. Further, (8) in (1) above, the sample sentence analysis unit has a designation means for designating a translated word of the second language corresponding to each word of the sample sentence, in addition to the dependency analysis. However, in the dictionary information storage section, when there is a possibility that there are a plurality of original forms for one solid word form, the dictionary information storage unit has a selection means for selecting an original form that matches a designated translated word, and further (9) the above In (1), the dictionary combining unit stores the dictionary information stored in the dictionary information storage unit according to a prescribed description format according to frequency information stored in the storage unit. (10) In (1) above, in the dictionary synthesizing unit in the above (1), the information specified by the dictionary information specifying unit and the translation word specified by the sample sentence analyzing unit are combined. If the pair is not in the general-purpose dictionary, it has a means for newly adding the pair as dictionary information,
Furthermore, (11) In (1) above, in addition to the information specified by the dictionary information specifying unit in the dictionary synthesizing unit, the priority of the information that has a high priority in the general-purpose dictionary Adjusting means to add to the dictionary information, and (12) above (1)
In, a translation word dictionary including a translation word, a part of speech, and a co-occurrence frame is created from a bilingual general-purpose dictionary, and a pair of the part of speech information identified by the dictionary information identification unit and the translation word specified by the user in the sample sentence analysis unit is used. , A corresponding co-occurring case frame is extracted from the translation dictionary and is added to the dictionary information. Less than,
A description will be given based on an embodiment of the present invention.

FIG. 1 is a block diagram for explaining an embodiment of a dictionary creating apparatus according to the present invention. In the figure, 1 is a sample text extracting section, 2 is an analysis information instructing section, 3 is an analysis result storing section, 4 is a dictionary information specifying unit, 5 is a dictionary information storage unit, 6 is a dictionary synthesizing unit, 7 is a sample file, 8 is a table file, 9 is a dictionary information file, 10 is a format file, 11 is a customized dictionary file, and 12 is a corpus. , 13 are analysis result files, 14 is a general-purpose dictionary file, and 15 is a translation word dictionary file.

The sample text extraction unit 1 extracts sample text of appropriate quantity and quality from the corpus 12 of the field to be processed. The extracted text is sent to the analysis information instruction unit 2 or stored in the sample file 7 as needed. In the analysis information instruction unit 2, the user instructs analysis information on the presented sample text. As the analysis method, a method suitable for the characteristics of the processing target language and the ability of the user, such as dependency analysis and phrase structure analysis, is adopted. Further, information can be referred to from the general-purpose dictionary file 14 as needed.
The analysis result is stored in the analysis result file 13 or directly sent to the next process after being subjected to an appropriate process in the analysis result storage unit 3. Further, at the time of analysis, the table file 8 storing the dependency pattern and the like is referred to.

The dictionary information specifying unit 4 specifies the information that matches the result of the sample analysis from the information obtained by searching the general-purpose dictionary. The specified information is collected in the dictionary information storage unit 5 in units of original forms. The collected information can be stored in the dictionary information file 9. The stored information is combined into a dictionary description by the dictionary combining unit 6 according to the format of the format file 10 and stored in the customized dictionary file 11. The format of the dictionary description may directly follow the format of the general-purpose dictionary file. Further, when the dictionaries are combined, the information in the general-purpose dictionary file 14 and the translated word dictionary file 16 is supplemented as necessary.

FIG. 2 is a diagram showing the structure of the table file shown in FIG. 1. Reference numeral 21 is a dependency pattern table,
Reference numeral 22 is a dependency matching table, 23 is a syntax rule table, 24 is a term structure match table, 25 is a priority order table, and 26 is an original form / variant form correspondence table. The modification pattern table 21 stores modification patterns used in sample sentence analysis. The dependency matching table 22 is for identifying information suitable for usage in the sample sentence from the search result of the general-purpose dictionary. The original form / variant form correspondence table 26 is for collecting the specified information in units of original forms. FIG. 3 is a diagram showing an example of a dependency analysis pattern table in the embodiment of the present invention. All words except head modify one word other than yourself in any of these relations.

FIG. 4 is a flowchart of dependency analysis in the sample sentence analysis unit. Hereinafter, each step will be described in order. Load the sample text (st
ep1), each sentence is divided into words (step 2), and a general dictionary is searched for each divided word (step 3). Each sentence is displayed to the user (step 4) and waits for the user's designation. For one word, the qualification destination and the relation name (step 5 to step 7) or the translation word (step 8) is designated as necessary. In the case of designating a translated word, the user can refer to the search result of the general-purpose dictionary as needed (step 13 to step 14). The specified information is stored together with the entry word form obtained by the general-purpose dictionary search (step 9 to step 10). If there is an instruction to cancel the designation on the way, the designation items after that are canceled and a new designation is waited for (step 15). On the way, sentence skip instruction (step1
If there is no 6) or end instruction (step 17), this is repeated for all the words in one sentence (step 11), and for all the sentences in the sample text (step 12).

FIGS. 5A and 5B are flowcharts of the dictionary information specifying process in the dictionary information specifying unit. Hereinafter, each step will be described in order. The stored information is read word by word (step 1), the dependency matching table 22 is called (step 2), and the part of speech that does not match the matching of the dependency and the dependency is deleted (steps 3 and 4). Next, the syntax rule table is called (step 5), and if there is a part of speech to which the rule is applied (step 6), the part of speech that meets the deletion condition is deleted (steps 7 and 8). This is repeated until no part of speech is applied (step 9). Next, the term structure match table is called (step 10), and if there is a part of speech described in the table (step 11), those that do not match the modified information and the term structure pattern are deleted (steps 12 and 13), and the part of speech to be checked is Repeat until no more (step 14). If the candidate part-of-speech cannot be narrowed down to one by the processing so far (step 1
5), the priority table is called (step 16), and if the modification relation and the candidate part of speech are described in the table (step 1)
(7, 18), delete parts of speech with lower priority (step 1
9). This is repeated until the last word (step 20). Figure 6
9 to 9 are diagrams showing examples of the dependency consistency table, the syntax rule table, the term structure match table, and the priority table, respectively. FIG. 10 is a diagram showing an example of the result of the dependency analysis in FIG. 4 and the dictionary information identification processing in FIG.

FIG. 11 is a flow chart of a process of collecting information in prototype units in the dictionary information storage unit.
Hereinafter, each step will be described in order. The specified information is read word by word (step 1), and when there are a plurality of original forms (step 2), the original forms that do not match the specified part of speech or the specified translated word are deleted (steps 11 and 13). If the original form and the solid form do not match (step 3), the word form is converted to the original form (step 4), and if the part of speech has a variant code (step 5), it is converted to the original form code (step 6), and the part of speech is duplicated. If occurs, delete it (step 7). If the word has already been registered in the dictionary information file (step 8), its frequency information and the like are updated (step 9), and if not registered, new information is created (step 14), and the process is repeated until the last word (step 14). step10). FIG. 12 shows the steps from step 11 to st.
It is a figure which shows an example of the result of the original form specific processing of ep13. In the upper part, the original form is specified by the specific part of speech, and in the lower part, the original form is specified by the correspondence with the designated translated word. FIG. 13 is a diagram showing an example of the original form / variant form correspondence table of FIG. 14
FIG. 13 is a diagram showing an example of a result obtained by collecting the analysis information identification results of FIG. 10 in prototype units by the processing of FIG. 12.

FIG. 18 is a flowchart of the dictionary synthesizing process in the dictionary synthesizing unit. Hereinafter, each step will be described in order. The sample information collected as shown in FIG. 14 is read word by word (step 1) and written in the customized dictionary file according to the format (step 1).
2). Complement the missing information from the general-purpose dictionary file (step
3) If necessary, manipulate the priority according to the number of appearances (step
4, 8). If a case frame requires a part of speech (step
5), copy the case frame from the translation dictionary (step
9). Information that is not included in the sample information but has a high priority in the general-purpose dictionary is written in the customized dictionary file by operating the priority if necessary (steps 6 and 10).
~ 12). This is repeated until the last word (step 13).

FIG. 15 is a diagram showing an example of information in a general-purpose dictionary file. FIG. 16 is a diagram showing an example of the format of the customized dictionary. FIG. 17 is a diagram showing an example of a translated word dictionary created from general-purpose dictionary information. Even when a translated word that is not in the general-purpose dictionary is added, it is possible to add an appropriate case frame by extracting the information of the translated word dictionary from the combination of the specified translated word and the identified part-of-speech.
FIG. 19 is a diagram showing an example of the information of the customized dictionary synthesized by the processing of FIG.

[0015]

As is apparent from the above description, the present invention has the following effects. (1) Effect corresponding to claim 1: According to the present invention, a sample text having an appropriate quality and quantity is extracted from a document to be processed, and a linguistic analysis of the sample text is performed. The customized dictionary data can be efficiently obtained without doing so. (2) Effect corresponding to claim 2: Especially as an analysis method,
The sample sentence is divided into words, and the dependency analysis is performed to specify the modification target word and its relation name for each word. By comparing the dependency consistency with the information in the general-purpose dictionary, the usage in the sample sentence is determined. By automatically specifying, it is possible to avoid selection mistakes of general-purpose dictionary information due to user subjectivity, and more accurate information can be efficiently obtained. (3) Effects corresponding to claims 3 to 5: When it is not possible to specify only by the consistency of dependency, the processing can be refined by using a syntactic rule, a term structure match, a priority order, or the like. (4) Effect corresponding to claim 6: Information that can be directly registered as dictionary data can be obtained by converting the information in solid form units obtained from the sample sentence into information in units of original forms and collecting them. (5) Effects corresponding to claims 7 and 8: When there is a possibility of plural original forms for one solid word form, the original form is obtained by the correspondence with the information specified by the above means and the translation specified by the user. It becomes possible to obtain accurate information by narrowing down to one. (6) Effects corresponding to claims 9 to 12: The information obtained by the means is combined in accordance with a predetermined format, and information not specified in the general dictionary but specified by the user is added, or a sample sentence is added. By appropriately supplementing the information that cannot be directly obtained from the general-purpose dictionary, it becomes possible to efficiently and precisely create a customized dictionary suitable for processing the target document.

[Brief description of drawings]

FIG. 1 is a configuration diagram for explaining an embodiment of a dictionary creation device according to the present invention.

FIG. 2 is a configuration diagram of a table file in FIG.

FIG. 3 is a diagram showing a dependency pattern table of the present invention.

FIG. 4 is a diagram showing a flowchart of dependency analysis in the sample sentence analysis unit of the present invention.

FIG. 5 is a diagram showing a flowchart of dictionary information specifying processing in the dictionary information specifying unit of the present invention.

FIG. 6 is a diagram showing a dependency matching table of the present invention.

FIG. 7 is a diagram showing a syntax rule table of the present invention.

FIG. 8 is a diagram showing a term structure match table of the present invention.

FIG. 9 is a diagram showing a priority table of the present invention.

FIG. 10 is a diagram showing an analysis information identification result of the present invention.

FIG. 11 is a diagram showing a flowchart of information gathering processing for each original form in the dictionary information storage unit of the present invention.

FIG. 12 is a diagram showing a result of the original shape specifying process of the present invention.

FIG. 13 is a diagram showing an original form / variant form correspondence table of the present invention.

FIG. 14 is a diagram showing sample information of the present invention.

FIG. 15 is a diagram showing general dictionary information of the present invention.

FIG. 16 is a diagram showing format information of the present invention.

FIG. 17 is a diagram showing a translated word dictionary of the present invention.

FIG. 18 is a diagram showing a flowchart of a dictionary synthesizing process in the dictionary synthesizing unit of the present invention.

FIG. 19 is a diagram showing information in a customized dictionary of the present invention.

[Explanation of Codes] 1 ... Sample text extraction unit, 2 ... Analysis information instruction unit, 3
... analysis result storage unit, 4 ... dictionary information specifying unit, 5 ... dictionary information storage unit, 6 ... dictionary combining unit, 7 ... sample file, 8 ...
Table file, 9 ... dictionary information file, 10 ... format file, 11 ... customized dictionary file,
12 ... Corpus, 13 ... Analysis result file, 14 ... General dictionary file, 15 ... Translation dictionary file.

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] November 19, 1993

[Procedure Amendment 1]

[Document name to be corrected] Drawing

[Correction target item name] All drawings

[Correction method] Change

[Correction content]

[Figure 7]

[Figure 1]

[Fig. 2]

[Figure 3]

[Figure 9]

FIG. 16

[Figure 4]

[Figure 5]

[Figure 8]

[Fig. 13]

FIG. 19

[Figure 6]

[Fig. 12]

[Figure 10]

FIG. 14

FIG. 11

FIG. 15

FIG. 17

FIG. 18

Claims (12)

[Claims] 1. A sample text extracting unit for extracting a sample text from a document of a processing target field, an analysis information instructing unit for presenting the extracted sample text and instructing analysis information of a sentence in the text, Based on the analysis result storage unit that stores the analysis information and the analysis result stored in the analysis result storage unit, for each word in the sample sentence, dictionary information that specifies information suitable for usage in the sample from information in the general-purpose dictionary Based on the specification unit, the information specified by the dictionary information specification unit, the dictionary information storage unit that stores frequency information such as the frequency of appearance of words, and the information of a general-purpose dictionary, if necessary, based on the prescribed description format. And a dictionary synthesizing unit for synthesizing the information stored in the dictionary information storage unit into a dictionary format. 2. The analysis information instructing unit, the information to be instructed is a relation of each word in a sample sentence and its qualified relation name, and based on the information, in the analysis result storage unit,
A qualification destination, a qualification source, and each qualification relation name are stored for each word, and the dictionary information specifying unit specifies information matching with the dependency consistency from the dictionary lookup information by the general-purpose dictionary for each word. The dictionary creating apparatus according to claim 1, further comprising a specifying unit for performing the setting. 3. The dictionary information identifying unit has an excluding means for excluding the inappropriate information by applying a syntactic rule when the inappropriate information in the general dictionary cannot be eliminated only by the dependency matching. The dictionary creation device according to claim 2, wherein 4. The dictionary information specifying unit specifies a term structure such as a verb in a sample sentence according to the number and types of terms extracted from the dependency information, and stores information matching the specified term structure in a general-purpose dictionary. The dictionary creating apparatus according to claim 2, further comprising an extracting unit that extracts the information. 5. If the dictionary information specifying unit cannot specify the information suitable for the usage in the sample sentence from the information of the general dictionary by the removing unit and the extracting unit, the dictionary information specifying unit arbitrarily determines according to the language property of the sample text. 5. The dictionary creation device according to claim 3, further comprising means for specifying information according to the priority order. 6. The dictionary information storage unit has means for collectively listing the dictionary information obtained in units of solid word forms in a sample sentence as information in units of original forms in a general-purpose dictionary. Dictionary creation device. 7. In the dictionary information storage unit, when there is a possibility of a plurality of original forms for one solid form, a selecting means for selecting an original form that matches the information specified by the dictionary information specifying unit is provided. The dictionary creation apparatus according to claim 1, wherein the dictionary creation apparatus comprises. 8. The sample sentence analysis unit has a designation unit for designating a translated word of a second language corresponding to each word of the sample sentence, in addition to the dependency analysis, and one of the dictionary information storage units 2. The dictionary creating apparatus according to claim 1, further comprising a selection unit for selecting an original form that matches a designated translation when there are a plurality of original forms for the solid form. 9. The synthesizing unit for synthesizing the dictionary information stored in the dictionary information storage unit in the dictionary synthesizing unit according to the frequency information stored in the storing unit according to a prescribed description format. The dictionary creating apparatus according to claim 1, further comprising means. 10. If the pair of the information specified by the dictionary information specifying unit and the translated word specified by the sample sentence analyzing unit is not in the general-purpose dictionary, the dictionary synthesizing unit newly creates the pair. The dictionary creating apparatus according to claim 1, further comprising means for adding as dictionary information. 11. The dictionary synthesizing unit, in addition to the information specified by the dictionary information specifying unit, adjusts the priority of information that has a high priority in a general-purpose dictionary and adjusts the dictionary. The dictionary creating apparatus according to claim 1, further comprising an adding unit that adds the information. 12. A translation word dictionary comprising a translation word, a part of speech, and a co-occurrence frame is created from a bilingual general-purpose dictionary, and a pair of the part of speech information specified by the dictionary information specifying unit and a translation word specified by the user in the sample sentence analysis unit. The dictionary creating apparatus according to claim 1, further comprising: an adding unit that extracts a corresponding co-occurrence frame from a translation dictionary by using, and adds it to the dictionary information.