JPH0785072A - Data base selector - Google Patents

Abstract

PURPOSE:To provide the data base selector which can automatically select a desired data base and is rich in versatility concerning the addition of a new data base. CONSTITUTION:Concerning each semantic attribute applied to each word provided at a character string analyzing means 2, a point for each data base is applied from a data base decision table preserved in a data base decision table preserving means 5 and the name of a data base, for which the result of summing up the points with respect to the respective data bases of all the semantic attributes provided at the character string analyzing means 2 is more than a prescribed point, is decided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a database selecting device, and more particularly to a database selecting device which automatically selects a database that meets a user's search request from a plurality of search target databases.

[0002]

2. Description of the Related Art A conventional database selecting apparatus selects a database that meets a search request from a plurality of databases to be searched, in advance, based on the knowledge of the field to be searched and the relationship between the databases. Defined as a structure, analyze the field based on knowledge,
A database associated with the field is selected by making a decision by inquiring the user (Japanese Patent Laid-Open No. 5-101101).

FIG. 13 shows the configuration of a conventional database selection device. The database selection device shown in the figure comprises search field knowledge 10, database knowledge 11, and a selection processing unit 12.

The search field knowledge 10 expresses a plurality of search fields for each predetermined field classification in a tree structure of a hierarchy set based on a predetermined relationship, and positions the search field at each node of the tree structure. Is the knowledge.

The database knowledge 11 is information in which a plurality of databases are associated with the search fields held in the search field knowledge 10. The information indicating the search field is represented, for example, in a state in which the required search fields are combined by a logical expression formed by the logical sum of the logical product terms.

The selection processing unit 12 uses the search field knowledge 1 described above.
A desired database is searched by performing the following operation using 0 and database knowledge 11.

First, as an initial state, all categories are used as candidate search fields, and all databases are used as candidate databases.

When the search field corresponding to the candidate database according to the database knowledge 11 corresponds to the candidate search field or the search field connected to the upper or lower part of the candidate search field in the tree structure, the candidate database and the candidate search field are Correspondingly, the following process is repeated until the candidate search field does not change by the following process.

If there is a candidate search field that does not correspond to the candidate database, the candidate search field is deleted and it corresponds to all candidate databases, or the candidate database does not change even if it is deleted as a selected one. If there is a search field, each candidate search field is deleted. If there is a search field located in a lower node in the tree structure, the search field when there is a lower node is added as a candidate search field.

The selection processing unit 12 displays the candidate search fields to the searcher, and regarding the displayed candidate search fields,
A searcher selects and inputs a desired candidate search field. Further, the candidate database that does not correspond to the candidate search field selected from the candidate database is deleted.

In the above processing, the candidate database is not deleted or the candidate database when a predetermined stop instruction is received is set as the selection processing result.

[0012]

However, in the above conventional technique, when a database of a different field is newly added, it is necessary to manually reconstruct the knowledge system (candidate search field). In addition, when the relationship between the field knowledge and the database is given from a viewpoint different from the existing relationship, the consistency with the existing relationship is lost and correct processing cannot be performed. There is a problem of poor versatility.

The present invention has been made in view of the above points, solves the conventional problems described above, can automatically select a database for a search request, and is highly versatile for adding a new database. An object is to provide a database selection device.

[0014]

FIG. 1 is a block diagram showing the principle of the present invention.

The present invention is a database selection device for automatically selecting a database matching a user's search request from a plurality of search target databases, and for inputting a character string relating to the database to be searched by the searcher. Of the character string input means 1, a character string analysis means 2 for extracting words from the character string input from the character string input means 1 and giving a semantic attribute to each extracted word, and a document in each database. A document analysis unit 3 that extracts words from a group, adds a semantic attribute to each extracted word, and counts the appearance frequency of each semantic attribute for each database, and each database counted by the document analysis unit 3. The score for each database for each semantic attribute is calculated from the frequency of occurrence of each semantic attribute for each database, and the score for each database for the semantic attribute and each semantic attribute is calculated. And a database decision table creating means 4 for creating a database decision table describing the name of the database, a database decision table saving means 5 for saving the database decision table created by the database decision table creating means 4, and a character string analyzing means 2 For each semantic attribute given to each word obtained in step 1, a score is given for each database by referring to the database determination table stored in the database determination table storage means 5, and the character string analysis means 2 The result of totaling the scores for each database of all the semantic attributes obtained in step 5 outputs the database determining means 6 for determining the name of the database having a predetermined score or more, and the name of the database determined by the database determining means 6. The output means 7 is provided.

[0016]

According to the present invention, the semantic attribute is directly extracted from the document group in the database, the table is created from the score based on the appearance frequency of the semantic attribute, and the database is selected using the table. Even if a new database is added, the score of the semantic attribute for that database is calculated and added to the table. Even if a new database is added, the score of the semantic attribute for that database is calculated and the table is calculated. In addition to the above, it is possible to select a database having high versatility with respect to addition of a new database.

[0017]

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

FIG. 2 shows the configuration of a database selection device according to an embodiment of the present invention. The database selection device shown in the figure includes a character string input unit 1, a character string analysis unit 2, a document analysis unit 3, a database (hereinafter, DB) determination table creation unit 4, a DB determination table storage unit 5, a DB determination unit 6, It is composed of an output unit 7, a word dictionary 8 and a DB determination table 9.

In the character string input unit 1, a user inputs a search request character string.

The character string analysis unit 2 divides the search request character string input from the character string input unit 1 into words and extracts the words. Further, a semantic attribute is added to each word extracted by the morphological analysis processing from the word dictionary, and the word to which the semantic attribute is added is output. Regarding the morphological analysis process, many methods have been proposed, and here,
The method does not matter.

The document analysis unit 3 extracts words from a plurality of documents randomly extracted from each DB from the document group extracted in the same manner as the character string analysis unit 2, performs morphological analysis, and extracts each word. On the other hand, a semantic attribute is added from the word dictionary 8 and the appearance frequency of the semantic attribute is counted for each DB. Specifically, the document analysis unit 3 uses a keyword automatic extraction process for extracting a word that is a keyword based on information such as the context of words and the appearance position, so that instead of word extraction by morphological analysis, A keyword word is extracted from a document in the DB, a keyword word is extracted, and the word dictionary 8 is added to the word extracted as the keyword.
A semantic attribute may be added according to the above, and the appearance frequency of the semantic attribute may be counted for each DB.
Many techniques have been proposed for the automatic keyword extraction process, and the details of the technique are not limited in the present invention.

FIG. 3 shows an example of a word dictionary according to an embodiment of the present invention. The word dictionary 8 is used by the character string analysis unit 2 and the document analysis unit 3 to add a semantic attribute. As shown in FIG. 3, the word dictionary 8 describes a part of speech and a semantic attribute (a plurality of descriptions are allowed for one word) for each word. For example, if the word is "phone,"
The part-of-speech is “syntax” and its semantic attributes are “telephone” and “device”.

The DB determination table creation unit 4 calculates a score for each DB for each meaning attribute from the appearance frequency of each meaning attribute counted by the document analysis unit 3 for each DB, and the meaning attribute and its meaning attribute. The DB determination table 9 describing the score for each DB and the name of each DB is created. The calculation of the score in the DB determination table creation unit 4 is
When a certain semantic attribute appears more frequently in a particular DB than the total number of semantic attributes that have appeared, the specific DB
The statistical processing is performed so that the score for

Regarding the statistical processing, many methods have been conventionally proposed in the field of statistics, and any of these methods may be used. Here, as one method of statistical processing, description will be given by the score for each DB of each semantic attribute calculated by the following equations (1) and (2).

[0025]

[Equation 1]

In the above equation, Y _jk is the score of the semantic attribute j in DB _k , F _jk is the frequency of the semantic attribute j in DB _k , and M _jk is the random appearance of the semantic attribute j without depending on the DB. Frequency (theoretical frequency) in DB _k when
m is the number of DBs, and n is the number of different semantic attributes.

According to the above equation, the score F is F _jk -M _jk >
When 0 (occurs at a frequency of appearing at random or more), it is high, and when F _jk −M _jk ≦ 0 (does not appear at a frequency of appearing at random), it becomes low. Calculated.

FIG. 4 shows an example of a DB decision table created by the DB decision table creating section according to an embodiment of the present invention. As shown in the figure, the DB determination table 9 describes the score of each DB for each semantic attribute. For example, when the semantic attribute is "mobile", the score according to the appearance frequency of the semantic attribute of the word "mobile" in DB1 is:
The value is "2.3", which means that they appear at a frequency higher than that of random appearance. Also, the score of DB2 is
It is “−0.1”, which means that it does not appear at the frequency of appearing randomly.

As described above, the DB determination table 9 is generated by the score according to the appearance frequency of the semantic attribute for each database.

The DB determination table storage unit 5 stores the DB determination table 9 created by the DB determination table creation unit 4. The DB decision table 9 is created by the DB decision table creation unit 4 when a change such as new addition or deletion is made to the target DB, and the DB decision table 9 stored in the DB decision table storage unit 5 Can be rewritten. Target D
When there is no change such as new addition or deletion for B, DB
The DB stored in the DB determination table storage unit 5 without the generation of the DB determination table 9 by the determination table creation unit 4
Use decision table.

The DB determination unit 6 uses the DB determination table 9 stored in the DB determination table storage unit 5 for each semantic attribute given to each word obtained by the character string analysis unit 2. , The score for each DB is added, and the character string analysis unit 2
The name of the DB having a score equal to or greater than the value designated in advance is output to the output unit 7 from the result of totaling the scores of all the semantic attributes obtained for each DB.

The output unit 7 receives the D sent from the DB determination unit 6.
The name of B is output.

The database selection process will be specifically described below. Here, the target DB is DB
1, DB2, DB3, and in each DB, FIG.
It is assumed that multiple documents such as those shown in are stored, and that "mobile phone" is input as the user search request character string.

FIG. 6 is a flow chart showing the operation of the database selection processing according to the embodiment of the present invention.

Step 1) The character string input unit 1 inputs a search request character string from the user and transfers it to the character string analysis unit 2.

Step 2) The character string analysis unit 2 divides the character string into words by using the word dictionary 8 shown in FIG. 3, assigns a semantic attribute to each word, and adds the semantic attribute to the word. Is output. FIG. 7 shows an analysis result in the character string analysis unit according to the embodiment of the present invention. As shown in the figure, by morphological analysis, the input character string "portable phone" is divided into "mobile", "use", "no", "telephone", ".",
Each has a semantic attribute of "mobile", "none", "none",
"Telephone", "Device" and "None" are added. Therefore,
The analysis result has the semantic attribute "mobile" for the extracted word "mobile" and the semantic attribute "telephone" for "phone".
It becomes "equipment". These analysis results are sent to the DB determining unit 6.

Step 3) The document analysis unit 3 extracts words from a document group consisting of a plurality of documents in each DB by the same morphological analysis process as the character string analysis unit 2, and adds a semantic attribute to each extracted word. Then, the appearance frequency of the semantic attribute is counted for each DB. FIG. 8 shows the result of the document analysis unit 3 when a word is extracted from the target DB by the morphological analysis process.
Shown in. FIG. 8 shows an analysis result (No. 1) of the document analysis unit according to the embodiment of the present invention. For example, as shown in FIG. 8, as a part of the documents randomly extracted from the document group of DB1 shown in FIG. 5, “Trend of mobile phone: Trend of mobile phone is (omitted)” “Portable "Phone (omitted)" (1 document in "") is extracted as a word to which a semantic attribute is attached by morphological analysis processing (abbreviated above is omitted), "mobile"
(2), "telephone" (3), "tendency" (2), "carry" (1) are obtained, and each semantic attribute "mobile" (3), "telephone" (3) ), "Equipment" (three),
"Formation" (2) and "Transport" (1) can be obtained. The number of semantic attributes obtained in this way is counted for all documents randomly extracted from DB1 and
By performing the same processing for, the appearance frequency of each semantic attribute for each DB can be obtained.

By performing the above-described processing, as shown in the figure, the DB stores a document group as shown in the figure a. Extract words from this document group,
A semantic attribute is given to each word divided by the morphological analysis process as shown in FIG. Next, the appearance frequency of the determined semantic attribute is counted. For example, DB1
The occurrence frequency of “transport” is 110 times, and “formation” of DB3 appears 80 times.

Further, by using the keyword automatic extraction process, the document analysis unit 3 extracts a word from the document group consisting of a plurality of documents in each DB by the keyword extraction process, and makes a meaning for each extracted word. Attribute is added to each D
The appearance frequency of the semantic attribute may be counted for each B. FIG. 9 shows an analysis result (No. 2) of the document analysis unit according to the embodiment of the present invention. This figure shows the result of the document analysis unit 3 when a word is extracted by the keyword automatic extraction processing.

As shown in the same figure, as a part of the documents randomly extracted from the document group of DB1 shown in FIG. 4, "Tendency of mobile phones: The tendency of mobile phones is (omitted)" If you extract a word that is a keyword (the above omitted part is omitted) by the keyword extraction process for a "portable phone (omitted)" (the one inside ""), the word not registered as a keyword (here, "trend") ,
“Mobile” is not extracted, and “mobile” (2) and “phone” (3) are obtained.
(2), "telephone" (3), and "device" (3) are obtained. The number of semantic attributes obtained in this way is stored in DB1.
By counting all the documents randomly extracted from, and performing the same process for other DB,
The appearance frequency of each semantic attribute for each DB is obtained. The result of the document analysis is transferred to the DB determination table creation unit 4.

Step 4) DB determination table creating section 4
Calculates the score for each DB of each semantic attribute from the result of the document analysis unit 3 by statistical processing, and creates the DB determination table 9. FIG. 4 shows an example of the DB determination table output from the DB determination table creation unit 4 for the target DB. Here, for the sake of explanation, as a result of the document analysis unit 3, a result when a word is extracted by a morphological analysis process, that is, a result of FIG. 8 is used. However, the result when a word is extracted by the keyword automatic extraction process, that is,
The result of FIG. 9 can be treated similarly. In addition, the calculation here is performed only for the five kinds of semantic attributes obtained from each DB shown in FIG. 8 (“mobile”, “transport”, “telephone”, “apparatus”, and “formation”). And

For example, in the case of the meaning attribute "mobile" shown in FIG. 8, first, the frequency M _{jk of} each meaning attribute with respect to each DB is obtained from the appearance frequency of each meaning attribute with respect to each DB, and Equation (2). This frequency M _jk and each DB of the semantic attribute "mobile"
Appearance frequency (150 in DB1, 10 in DB2
5 and 38 in DB3) is applied to the expression (1), and the score in each DB of the semantic attribute "mobile" (2.3, D in DB1)
In B2, -0.1 is obtained, and in DB3, -3.6) is obtained.

Step 5) DB determination table storage unit 5
Saves the DB determination table 9 created by the DB determination table creation unit 4.

Step 6) The DB decision unit 6 inputs the result of the character string analysis unit 2 and the DB stored in the DB decision table storage unit 5 for the inputted semantic attribute of each word.
Using the determination table 9, a score is given to each DB, the scores for all DBs of all the inputted semantic attributes are summed up for each DB, and the largest score having a positive value among the scores is given. The score of each DB having a positive value is divided and the name of the DB having a score of a predetermined value or more is output. Figure 10
The result of a DB determination part with respect to the example of the input character string of one Example of this invention is shown. In the case of the example of the input character string, “mobile”, “telephone”, and “device” are input as the semantic attributes, and “mobile”: (2. 3, -0.1, -3.6) "Telephone": (3.8, 2.2, -27.7) "Device": (-1.7, 0.0, 4.6) Is obtained. If these are summed up for each DB, D
The total score of B1 is 4.4, the total score of DB2 is 2.1, DB
A total score of 3 is −26.7. Of these, each D is the score of DB1 having the largest positive score 4.4.
Divide B's score. As a result, DB1 is 1.00,
If the result of DB2 is 0.48 and the pre-specified value is 0.5, DB with a value greater than 0.5
1 is output as a result.

Step 7) The output unit 7 outputs the name of DB1 determined by the DB determination unit 6.

Next, the processing when the DB 4 is added as a new DB will be described.

From the document group in the new DB 4, the document analysis unit 3
Then, a word is extracted by the morphological analysis process, a semantic attribute is given to each extracted word, the appearance frequency of each semantic attribute is added to the appearance frequency of each semantic attribute shown in FIG. The table creation unit 4 calculates the score and newly creates the DB determination table 9.

FIG. 11 shows an example of a DB determination table after adding a new DB according to an embodiment of the present invention. For example, the frequency of appearance of each semantic attribute in the DB4 obtained from the document group in the new DB4 by the document analysis unit 3 "mobile": 120, "transport":
90, "Telephone": 110, "Device": 60, "Form":
20 is added to the appearance frequency shown in FIG. 9, and in the DB determination table creation unit 4, the appearance frequency for each DB of each semantic attribute of the whole and the frequency M _jk for each DB of each semantic attribute are calculated from Equation (2). Required, this frequency M _jk and each DB of each semantic attribute
The appearance frequency in is applied to the equation (1) to obtain the score in each DB of each semantic attribute. In the case of the semantic attribute "mobile", the appearance frequency for each DB (150 for DB1, DB2
Frequency 105, DB3 38, DB4 120), and frequency M
_The score (0.5 for DB1, -0.10 for DB2, -5.4 for DB3, 3.7 for DB4) for each DB is obtained from _jk .

For the example of the input character string, the DB determination unit 6
12 shows the result of processing using the DB determination table after the addition of the new DB shown in FIG. In this case, “mobile”, “telephone”, and “device” are input as the semantic attributes, and “mobile”: (0.5, −1.0 for each DB using the DB determination table shown in FIG. 11). , -5.4, 3.
7), "Telephone": (2.4, 1.3, -29.3, 0.
8), "device": (-0.6, 0.3, 6.7, -2.
The score of 6) is obtained. For each of these DBs,
DB1 is 2.3, DB2 is 0.6, and DB3 is -28.
0, DB4 has a value of 1.9, and the DB having a positive value is divided by the largest positive value 2.3 (DB1), and DB1 has a value of 1.00 and DB2 has a value of 0. 26, DB
4 gets 0.83. Here, when the value specified in advance is 0.5, as a result, DB larger than 0.5
1 and DB4 are output. The output unit 7 outputs the DB names of DB1 and DB4.

As described above, according to the present invention, when a user's search request character string is input, a DB created from the information in each DB for a plurality of search target DBs from the character string.
A matching DB is selected using a frequency table (DB determination table) as the knowledge for determination. That is, by using the semantic attribute added to the words of the document group of the DB, the score is given according to the appearance frequency of the semantic attribute, and the DB having a high score for the semantic attribute obtained from the search request character string is obtained as a result. By outputting, even if a new DB is added,
By adding the frequency information in the new DB to the existing knowledge (DB decision table) and calculating numerical data,
It is possible to select a DB that is a search request target and can be used universally for adding a new DB.

[0051]

As described above, according to the present invention, a new DB in a different field is added by directly extracting the semantic attribute from the data in the DB and creating a table from the scores according to the appearance frequency of the attribute. Even if it is done, by calculating the score of the semantic attribute for the DB and adding it to the table, it is possible to add a versatile DB to the addition of a new DB.
Can be selected.

[Brief description of drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is a configuration diagram of a database selection device according to an embodiment of the present invention.

FIG. 3 is a diagram showing an example of a word dictionary according to an embodiment of the present invention.

FIG. 4 is a diagram showing an example of a DB determination table created by a DB determination table creation unit according to an embodiment of the present invention.

FIG. 5 is a diagram showing an example of a document in a database according to an embodiment of the present invention.

FIG. 6 is a flowchart showing an operation of a database selection process according to an embodiment of the present invention.

FIG. 7 is a diagram showing an analysis result in a character string analysis unit according to an embodiment of the present invention.

FIG. 8 is a diagram (part 1) showing an analysis result of the document analysis unit according to the embodiment of the present invention.

FIG. 9 is a diagram (No. 2) showing the analysis result of the document analysis unit according to the embodiment of the present invention.

FIG. 10 is a DB for input character strings according to an embodiment of the present invention.
It is a figure which shows the result of a determination part.

FIG. 11 is a diagram showing an example of a DB determination table after adding a new DB according to an embodiment of the present invention.

FIG. 12 is a diagram showing a result of a DB determination unit after adding a new DB according to an embodiment of the present invention.

FIG. 13 is a configuration diagram of a conventional database selection device.

[Explanation of symbols]

1 character string input means, character string input part 2 character string analysis means, character string analysis part 3 document analysis means, document analysis part 4 database decision table creation means, database decision table creation part 5 database decision table storage means, database decision table Storage unit 6 Database determination unit, database determination unit 7 Output unit, output unit 8 Word dictionary 9 Database determination table 10 Search field knowledge 11 Selection processing unit 12 Database knowledge

Claims (1)

[Claims] 1. A database selection device for automatically selecting a database matching a user's search request from a plurality of search target databases, the character being used by a searcher to input a character string relating to the database to be searched. A column input means, a character string analysis means for extracting a word from the character string input from the character string input means, and giving a semantic attribute to each extracted word, and a word from a document group in each database. A document analysis unit that extracts and assigns a semantic attribute to each extracted word, and counts the appearance frequency of each semantic attribute for each database, and each semantic attribute for each database counted by the document analysis unit The score for each database for each semantic attribute is calculated from the appearance frequency of the, and the score and data for each database of the semantic attribute and the semantic attribute are calculated. Database decision table creating means for creating a database decision table describing the name of the database, database decision table saving means for saving the database decision table created by the database decision table creating means, and character string analysis means For each meaning attribute given to each word given, a score is given for each database by referring to the database determination table stored in the database determination table storage means, and the character string analysis means The output of outputting the names of the databases determined by the database determination means for determining the names of the databases whose total score of all the obtained semantic attributes for each database is a predetermined score or more Day characterized by having means Database selection device.