JPH09198400A - Information retrieval device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To more precisely give priority to a retrieval result even if a keyword is a sentence and a clause and to improve retrieval efficiency by means of urging retrieval in order from the result of the highest priority. SOLUTION: A thesaurus development part 2 refers to a thesaurus 3 and thesaurus-develops the keyword inputted from an input part 1 and gives it to a retrieval part 4. The retrieval part 4 retrieves a text information storage part 5 by a word which is thesaurus-developed and gives the retrieval result to a thesaurus development check part 6, a relation check part 7 and a related word check part 8. The check parts 6, 7a and 8 check the presence or absence of thesaurus development, the matching of relation between a word with the keyword and the words, and the number and the positions of the keywords and related words, and give the check results to a priority calculation part 9. The priority calculation part 9 calculates priority with weighting on the check results. A display part 10 displays the retrieval result in accordance with the priority.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information retrieval apparatus which retrieves text information or the like using a keyword, calculates the degree of importance of each retrieval result, and displays the retrieval result according to the degree of importance. .

[0002]

2. Description of the Related Art In recent years, as the amount of documents that can be handled has increased due to the higher functionality of document processing, the amount of information to be searched tends to increase. The demand for gain is increasing. For example, regarding a device that searches for text information using keywords, various levels of information are mixed in the search results, from important information to garbage information that has little to do with keywords. The challenge is how to efficiently obtain important information from inside.

In response to such a request, a method of efficiently obtaining important information by prioritizing search results by keywords to notify the user and checking from the highest priority is practical. Has been converted.

By the way, some of the devices that perform information retrieval using keywords, such as the document retrieval device described in Japanese Patent Laid-Open No. 62-248032, have a thesaurus dictionary to minimize the omission of retrieval. There are not a few that have adopted the method of searching by expanding keywords with synonyms.

However, in this type of document retrieval apparatus, as a result of inflating a keyword with a synonym or the like in a thesaurus dictionary, the omission of retrieval can be prevented, while the retrieval result is further increased. Is more important than any other device.

Therefore, as a conventional representative example of prioritizing search results, an information search method described in Japanese Patent Laid-Open No. 59-223865 and a document search apparatus described in Japanese Patent Laid-Open No. 4-281565. Etc. are known. In the former search method, prioritization is performed according to the number of keywords included in the text to be searched, and in the latter device, priority is given based on the number and position (title, abstract, body) of keywords. It is carried out.

In addition, as a conventional technique for prioritizing search results, a method of calculating the priority of search results based on the number and position of keywords (words) and related words (words having a strong relationship with keywords), A method that extracts word-word relationships from keywords (words, clauses, sentences), and preferentially displays words that match the word-word relationships extracted from the search results. Is also known.

[0008]

The importance of a search result can be judged by the amount of description of a keyword. As the description of this keyword, a direct description that directly refers to a keyword and an indirect description that describes an item related to the keyword are given. A description is possible.

A method of prioritizing according to the number or position of keywords in search results (Japanese Patent Laid-Open No. 59-22386)
No. 5, or Japanese Patent Laid-Open No. 4-281565) only considers the direct description of the keywords. Especially when the number of sentences is small and the number of keywords is small, there is not much difference and the prioritization is performed with good accuracy. I couldn't.

In this respect, in the method of prioritizing according to the number and position of keywords and related words in the search result,
Since the keyword itself corresponds to the direct description and the related word corresponds to the indirect description, it can be said that the direct description and the indirect description are considered to some extent. With this method, even if the number of sentences is small, it is possible to prioritize accurately, but on the other hand, the keywords are limited to words, and when the keywords are sentences or clauses, prioritization itself cannot be performed. .

Therefore, as a method that enables prioritization even when a keyword is a sentence or a clause, a plurality of words and the relationship between these words are extracted from the keyword, and the same is obtained from the search results. Attention has been focused on a method of extracting a word and a relationship between the words and preferentially displaying the one (search result) in which the relationship between the words matches.

In this system, for example, when the keyword "create a document" is given, "document", "create", and the relation "wo case" between these words are extracted, and the search result Among them, when the relationship between words is the same, for example, "Create document", "Create document", "Create document", etc. are displayed with priority, the result is that the keyword is limited to the word The scope of application is dramatically increased.

However, according to this method, only the words in the keyword and the search result are evaluated as to whether or not the relationships between the words match. Therefore, the calculated priority is binary (match, It has stopped at (mismatch). This allows
Only rough prioritization is still possible, search results are displayed in order of priority, and checking is performed from the highest priority, and the expected effect cannot be expected from the viewpoint of efficiently obtaining important information. It was

The present invention has been made in order to solve the above-mentioned problems, and it is possible to extend the prioritization of search results by keywords so that they can be performed even when the keywords are sentences or clauses. We provide an information retrieval device that enables highly accurate prioritization of relevant search results through comprehensive evaluation based on various item checks, and prompts checks from the highest priority to obtain important information more efficiently. The purpose is to do.

[0015]

In order to achieve the above object, the first aspect of the present invention is such that text information storage means for storing text information, input means for inputting a search key, and input from the input means. From the search key, a search expression creating means for creating a search expression including a synonym that is a synonym expansion of the word of the search key, a related word storage means for storing a related word, and the search from the related word storage means. A related word acquisition unit that acquires a related word of the key, a search unit that searches the text information storage unit by the search formula, and a number of words that are synonym-expanded in the search formula for each search result of the search unit. And a synonym expansion checking means for checking whether the relation between words in each search result matches the relation between words in the search key. Check hand A related word check means for checking the number and position of the search key and its related words included in each search result for each search result, the synonym expansion check means, the relationship check means, and the related word check means. And a display unit for displaying each search result in accordance with the priority corresponding to each search result. To do.

A second aspect of the invention is a text information storage means for storing text information, an analysis information storage means for storing analysis information obtained by previously analyzing the text information, an input means for inputting a search key, and From the search key input from the input means, a search expression creating means for creating a search expression including a synonym obtained by expanding a synonym of the word of the search key, a related word storage means for storing a related word, and the related word A related word acquisition unit that acquires a related word of the search key from a storage unit, a search unit that searches the analysis information storage unit by the search formula, and a synonym in the search formula for each search result of the search unit. Synonym expansion checking means for checking how many expanded words are included, and whether the relationship between words in each search result matches the relationship between words in the search key for each search result. or not Relationship checking means for checking, related word checking means for checking the number and position of the search key and its related words included in each search result for each search result, the synonym expansion checking means, the relationship checking means, And a display unit for displaying each search result in accordance with the priority corresponding to each search result, based on each check result of the related word checking unit. It is characterized by doing.

[0017]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. FIG. 1 shows a schematic configuration of an information search device according to an embodiment of the first invention. This information search device includes an input unit 1, a thesaurus expansion unit 2, a thesaurus dictionary 3, a search unit 4, a text information storage unit 5, a thesaurus expansion check unit 6, a relationship check unit 7, a related word check unit 8, and a priority calculation unit. 9, display unit 10
It comprises.

The input unit 1 has a morphological analysis dictionary. When the keyword D1 is obtained from the user, the input unit 1 performs morphological analysis using the dictionary and transmits the analysis information D2 of the keyword D1 to the thesaurus expansion unit 2.

When the analysis information D2 of the keyword D1 is obtained, the thesaurus expansion unit 2 searches the thesaurus dictionary 3 for the noun and Sa verb of the keyword D1 to obtain synonym and related word information D3. Then, this synonym and related word information D
3, the keyword D1 is expanded into a synonym, and the resulting synonym expansion information and related word information D4 are transmitted to the search unit 4.

The thesaurus dictionary 3 stores information on synonyms and related words (words having a strong relationship) with respect to a headword. Here, a synonym is a word having the same meaning as a headword, and a related word is a word that has a different meaning but a stronger relationship with the headword. For example, regarding the entry word "AI, OA", AI ... synonyms "artificial intelligence ..." Related words "computer, expert system ..." OA ... synonyms "office automation ..." Related words "word processor, personal computer, fax ..." Information is stored.

When the retrieval section 4 obtains both the synonym expansion information and the related word information D4 of the keyword D1, the keyword D is obtained.
The text information storage unit 5 is searched by using a search expression treated as a product set for the independent words of 1 and a union set for the synonyms, and the search result D5 is searched by the thesaurus expansion unit 6, the relation checking unit 7,
The information is transmitted to the related word check unit 8.

Text information is stored in the text information storage section 5, and by searching for an arbitrary word,
You can get a description of the word.

The thesaurus expansion check unit 6 includes a search unit 4
When the search result D5 is obtained from the above, the synonym information obtained by the thesaurus expansion is referred to, and in the search result D5, it is checked how many words in the synonym are hit, and the check result D6 is obtained. Is transmitted to the priority calculation unit 9.

When the relation checking unit 7 obtains the retrieval result D5 from the retrieval unit 4, the relation checking unit 7 extracts relations between the keywords D1 and some of the words constituting the keyword D1. For each of the above search results D5, the relationship between the words is extracted and both (keyword D1
And the search result D5) to check whether the relation between the words matches, and the check result D7 is transmitted to the priority calculation unit 9.

When the related word check unit 8 obtains the search result D5 from the search unit 4, the related word check unit 8 checks the number and position of the keyword D1 and its related words existing in the search result D5, and the check result D8. Is transmitted to the priority calculation unit 9.

The priority calculation unit 9 calculates the search result D5,
When the thesaurus expansion check result D6, the relation check result D7, and the related word check result D8 are obtained, the priority of each of the search results D5 is calculated by reflecting these check results, and the result is calculated. The obtained priority information D9 is transmitted to the display unit 10. When the display unit 10 obtains the priority information D9 regarding the search result D5 from the priority calculation unit 9, the display unit 10 displays the corresponding search results D5 according to the priority.

Next, the operation of each component will be described using an actual example. FIG. 2 is a diagram in which an example of input and output of each component when the user specifies “create a document” as a keyword is also shown in the flow of processing.

When the input section 1 obtains the sentence "create a document" as the keyword D1, the morpheme analysis is performed, the keyword D1 is decomposed into morphemes, and the information D2 document "name" is added to this morpheme. ,, /, creation [support]] to the thesaurus expansion unit 2. here[],
/ Is part-of-speech information, a name is a noun, a / is a particle, and a verb is a verb. The keyword D1 is not limited to words, and can correspond to clauses and sentences, as can be seen from the above-mentioned example "create a document".

In the thesaurus expansion unit 2, the keyword D1
Analysis information D2 (Document "name", /, create [support])
Then, the thesaurus 3 is searched for the noun of the keyword D1 and the sa verb (document, creation), and the keyword D
Both synonym information 1 (synonyms “document, specification” of “document”, synonym “production” of “creation”) and related word information (related word “word processing” of “document creation”) D3 obtain.

Next, both synonymous and related information D3
The keyword D1 is expanded by using the synonym, and the result is synonym expansion information [(document, document, specification)
(Creation, production)] and related word information (word processor) information D4 are transmitted to the search unit 4. Where the thesaurus dictionary 3
The headword of the related word information of is a compound word in which the independent word of the keyword D1 is combined so that the keyword D1 can correspond to a sentence or a clause. In the case of this example, the keyword D1 "create a document" to a compound word (document creation)
Is searched for related word information. Also, in this example,
Although the related word information is obtained from the thesaurus dictionary 3, it may be directly input by the user or extracted on the spot from the text information. In the latter case, the related word information can be extracted from the text information by using the co-occurrence frequency with the keyword (frequency that appears near the keyword). For example, assuming that the range of co-occurrence is one sentence, when the keyword D1 is a word, the frequency of words co-occurring with the keyword D1 (included in the same sentence as the keyword D1) is counted, and unnecessary words are extracted from high-frequency words. When the keyword D1 is a sentence or a clause, the independent word is extracted from the keyword D1 and co-occurred with all the independent words of the keyword D1 (included in the sentence including all the independent words of the keyword D1). Count the frequency and remove unnecessary words from high-frequency words to make them related words.

When the retrieval unit 4 obtains both synonym expansion information [(document, document, specification) (create, produce)] and related word information (word processor) information D4 of the keyword D1, the keyword D1 becomes independent. The text information storage unit 5 is searched by using a search expression of a product set for terms and a union for synonyms, that is, [(document, document, specification) & (creation, production)].

In this example, as the search result D5, as shown in the column a of FIG. 2, [... create document by word processor ..., ... create specification ..., ... create document ..., create table by document. ...] is shown when four documents are searched. Here, ~ represents a document number. The search result D5 is transmitted from the search unit 4 to the thesaurus expansion check unit 6, the relation check unit 7, and the related word check unit 8, respectively.

When the thesaurus expansion check section 6 obtains the search result D5 (... create document by word processor ..., create specification document ..., create document, ... create table by document ...), the keyword D1 is set. It checks how many words are hit in the synonyms (documents, specifications, productions) obtained by expanding the thesaurus, and transmits the check result D6 to the priority calculation unit 9.

In this example, since only the document is hit with the synonym "specification", the check result D6 is obtained such that only one document is hit. In the column b of FIG. 2, this state is represented by a number "1" associated with the document. According to the present embodiment, the number of hit words in the thesaurus expanded word is output. However, when the thesaurus expanded word is hit, the meaning of the original word and the expanded word is output. It is also possible to calculate and output a specific distance and use it for prioritization.

When the relationship check unit 7 obtains the search result D5 (... Create document by word processor ..., Create specification document ..., Create document, ... Create table with document ...) from the search unit 4, the keyword D1 is obtained. While extracting the word-to-word relationship from among the above, the search result D5, that is, the document-to-word relationship is also extracted for each of the documents .about., And whether or not this relationship matches the above-mentioned keyword D1 relationship. Check the relationship about.

In this example, from the keyword D1 "create a document", (document ← [case] ← create) is extracted.
On the other hand, in the search result D5 above, the document "Create a document with a word processor" (Document ← [Case] ← Create), but the document "Specification create" (Specification ← [Case] ← Create )
However, from the document “document creation” (document ← [was case] ← created), and from the document “create a table with a document” (document ← [was case] ← created). Where [the case]
Indicates a relationship, and (document ← [corresponds] ← create) indicates that the [create case] of [create] is “document”.

In this case, for the relationship between the words extracted from the keyword D1 and the relationship between the words (document ← [was] ← created), it was extracted from the document "create a document with a word processor" (document ←
[Case] ← Create) matches and is extracted from the document “Creation of specifications” (Specification ← [Case] ← Create) matches and is extracted from the document “Creation of documents” (Document ← [Case ] ← Create)
Match, and the document extracted from the document “create table with document” (document ← [in case] ← created) does not match, and the relationship check unit 7 responds to these determinations to the priority calculation unit 9. The check result D7 is transmitted. This state is shown in FIG.
In column b, the documents are shown to have a relationship match, the documents, and the circles are associated with the documents.

When the related word check unit 8 obtains the search result D5 (... Create document by word processor ..., Create specification ..., Create document, ... Create table by document ...) from the search unit 4, The number and position (title, body) of the related words with respect to the keyword D1 in each document are checked (in this example, only the number of related words is output for simplification), and the check result D8 is given priority. It is transmitted to the degree calculation unit 9.

In this example, since the related word "word processor" of the above-mentioned keyword D1 "create a document" exists in the description of the document (creates a document by the word processor), only one related word exists in the document. Check result D8
Is transmitted to the priority calculation unit 9. About this situation,
In the column b of FIG. 2, a document in which a related word exists is associated with a number “1” indicating the number of the related words.

The priority calculation section 9 determines the search result D5 (
… Writing a document with a word processor …… Creating a specification…
Document creation, ... Creating a table with documents ...), the thesaurus expansion check result D6 (one document 2 hits), the relationship check result D7 (document, document, document match), the related word check result D8 ( Document is related word 1
(Including the number of), and these check results D6, D
Referring to 7 and D8, the document which is the above retrieval result D5 ~
Each priority is calculated.

The method of calculating the priority is performed, for example, by weighting according to each check item, and the specific weight value can be determined as follows, for example.

(1) When the synonym expanded by the thesaurus is hit, -500 per one (2) When the relationship between the words for the keyword and the word match, 2000 (3) 100 per one related word When the priority is calculated using the calculation result, the calculation result is as shown in Table 1 below (corresponding to column c in FIG. 2). That is, the document has a total of 2100 because the relation matches (2000) and there is one related word (100), and the document has a thesaurus expansion hit (-500) and the relation matches (2000). The total is 1500, which is (2000) because the documents only match the relationship, and (0) because the document does not meet the judgment requirements for any of the above check items.

[0043] Priority information D9 of the search result D5 (Document [2100] ...
Create a document with a word processor ..., Document [1500] ... Create specifications ..., Document [2000] ... Create document ..., Document [0] ... Create table with document ...) from the priority calculation unit 9 to the display unit 10. Be transmitted to.

The display unit 10 displays priority information D9 (document [2100] ... Create document by word processor ...) of the search result D5.
Document [1500] ... Create specifications ..., Document [2000] ... Create documents ..., Document [0] ... Create table with documents ...)
The search result D10 is displayed according to the priority among them. Specifically, as shown in the column d in FIG. 2, the priority information D9
Refer to the points in the above, and in descending order of the points, 1: Create a document with a word processor, 2: Create a specification document, 3: Create a document, 4: Create a table with a document ... Is displayed.

In this example, although the search results are sorted and displayed in order of priority, various modifications such as displaying only those having a certain score or more or displaying them in groups according to the score are of course possible. It is possible. The priority score may be displayed to the user. As described above, in the first invention, regarding the information search result using the keyword, the semantic distance from the keyword, the match / mismatch of the relationship between the words with respect to the keyword, the amount of description about the keyword, etc. It is possible to prioritize the information retrieval results with high accuracy by performing a quantitative evaluation, and it is possible to efficiently obtain important information by checking the information retrieval results with the highest priority.

Next, an embodiment of the second invention will be described. FIG. 3 shows a schematic configuration of an information retrieval apparatus according to the second invention, and the same reference numerals are given to the portions having the same functions as the respective portions of the apparatus according to the first invention in FIG. The apparatus according to the second aspect of the invention has a configuration in which the text information storage section 5 of the apparatus according to the first aspect of the invention is replaced with an analysis information storage section 11, and the text information storage section 5 is directly connected to the display section 10. is there.

According to the first aspect of the present invention, the search unit 4 targets the text information itself as a search target, and simply searches the text information including the search word based on the search word obtained by expanding the keyword D1 in the thesaurus. there were. Therefore, in calculating the priority of the search result D5 thereafter, in the relation check unit 7 or the related word check unit 8, the relation between the words for each of the search results D5 and the relation between the words and the related word with the keyword D1. It was necessary to analyze the search result D5 again in connection with the extraction of the number and the position of the search results.

The second invention is made in order to suppress an increase in search time due to this kind of analysis processing.
While the device according to the first invention searches the text information itself to calculate the priority, the device according to the second invention uses the analysis information obtained by analyzing the text in advance as the search target. The priority is calculated based on the search result of.

That is, in the device according to the second invention,
The text information stored in the text information storage unit 5 is analyzed in advance, and the result is stored in the analysis information storage unit 11 as analysis information. As the analysis information, at least divided morpheme information and part-of-speech information may be stored. This analysis information storage unit 11 is searched by the search unit 4
By searching for an arbitrary word, the analysis information of the description relating to that word can be directly obtained, and the load of analysis processing in the subsequent relation checking unit 7 and related word checking unit 8 can be reduced.

Next, an outline of the information retrieval operation of the device according to the second invention will be described. In the information search operation of this device, the search section 4 obtains a search word based on the keyword D1 as in the first aspect of the invention.

That is, when the retrieval unit 4 obtains both the synonym expansion information and the related word information D4 of the keyword D1, the analysis information storage unit uses the intersection between the independent words of the keyword D1 and the union between the synonyms. 11 is searched, and the search result D50 is transmitted to the thesaurus expansion check unit 6, the relation check unit 7, and the related word check unit 8, respectively.

Here, the analysis information storage unit 11 stores analysis information obtained by analyzing text information in advance,
By performing a search from the search unit 4 using a search expression that satisfies the above logical expression between the independent word and the synonym of the keyword D1, analysis information including a description regarding the search expression can be obtained as the search result D50. .

When the thesaurus expansion check unit 6, the relation check unit 7, and the related word check unit 8 obtain the search result D50 from the search unit 4, the semantic distance between the search result and the keyword D1 and the keyword D1 are obtained. A check process is performed for the matching of the word and the relation, the description amount of the keyword D1 and the like.

That is, the thesaurus expansion check unit 6
When the retrieval result D50 is obtained from the retrieval unit 4, refers to the synonym information obtained by the thesaurus expansion, and retrieves the retrieval result D50.
In 50, it is checked how many of the synonyms hit, and the check result D6 is transmitted to the priority calculation unit 9.

When the search result D50 is obtained from the search unit 4, the relation checking unit 7 extracts the keywords D1 from the words constituting the keyword D1 and the relation between these words, and retrieves the results D50. It is checked whether or not the relationship between the words in each of the search results and the relationship between these words match the relationship of the keyword D1 and the check result D7 is transmitted to the priority calculation unit 9. Here, the search result D50 is read from the analysis information obtained by analyzing the text information in advance, and it is not necessary to analyze the search result D50 again when the relationship check unit 7 checks the relationship. .

When the related word check unit 8 obtains the search result D50 from the search unit 4, the related word check unit 8 checks the number and position of the words matching the keyword D1 existing in the search result D50 and the related words. The check result D8 is transmitted to the priority calculation unit 9. Also here, since the search result D50 is one of the pieces of analysis information obtained by analyzing the text information in advance, it is not necessary to analyze the search result D50 again when the related word check unit 8 checks the related words. Become.

The priority calculation unit 9 calculates the search result D50,
When the thesaurus expansion check result D6, the relation check result D7, and the related word check result D8 are obtained, for each of these check results D6, D7, D8,
The priority of each search result D50 is calculated by performing weighting with a predetermined value for each check item,
The priority information D9 as the calculation result is transmitted to the display unit 10.

When the display unit 10 obtains the priority information D9 of the search result D50, it reads the text information corresponding to each search result D50 from the text information storage unit 5, and at the same time, reads the read text information. It is displayed to the user according to the priority given to the search result D50 corresponding to.

According to the second invention, the analysis information storage unit 11 stores the analysis information obtained by analyzing the text information in advance in the form of at least divided morpheme information and part-of-speech information. Since the search unit 4 is basically based on the process of using the keyword D1 to obtain the analysis information of the description corresponding to the keyword D1 as the search result D50, the search result is transmitted from the priority calculation unit 9 to the display unit 10. The priority information D9 includes at least divided morpheme information and part-of-speech information related to the keyword D1. Therefore, when displaying the search results in the order of priority on the display unit 10, the part-of-speech information is selected from the analysis information in the priority information D9 regardless of the method of searching the text information from the text information storage unit 5. Except for this, it may be possible to combine the original text information from the morpheme information.

Next, the operation of each component will be described using an actual example. FIG. 4 is a diagram in which an example of input and output of each component when the user specifies "create a document" as the keyword D1 is also shown in the process flow. In this case, the operation until the retrieval word is obtained by thesaurus expansion for the keyword D1 "create document" is the same as in the first aspect of the invention, and therefore the following processing will be described.

The search unit 4 obtains both synonym expansion information [(document, document, specification) (create, produce)] of the keyword D1 "create document" and related word information (word processor) D4. And the keyword D1 "Create a document"
Independent words in a product set, and synonyms in a union set are treated as a search expression, that is, [(document, document, specification) & (create,
[Production)] to search the analysis information storage unit 11, and as a result of the search, ... With word processor [name] / create document [name] / create [support] ...… / Create [name] of specification [name] ... ... document [name] creation [service name] ... ... analysis information (see column a in Fig. 4) of document [name] / table [name] / create [service]. The search result D50 is transmitted from the search unit 4 to the thesaurus expansion check unit 6, the relation check unit 7, and the related word check unit 8, respectively. The symbols (1) to (5) in the search result D50 represent document (content is analysis information) numbers.

The thesaurus expansion check unit 6 uses the search result D50 (... word processor [name] / create document [name] / create [service] ..., ... / spec [name] / create [service] ...,
... [document [name] create [service name] ..., document [name] / table [name] / create [service] ...), the synonym (the synonym obtained by thesaurus expansion of the above keyword D1) Documents, specifications, production), how many words are hit is checked, and the check result D6 is transmitted to the priority calculation unit 9.

In this example, as in the first embodiment, only the document is hit with the synonym "specification", so the check result D6 is such that only one document is hit. obtain. In the column b in FIG. 4, this state is represented by a number "1" associated with the document. Even in this case, in addition to the method of outputting how many words are hit in the thesaurus-expanded words, if the thesaurus-expanded words are hit, the semantic distance between the original word and the expanded words May be calculated and output to be used for prioritization.

The relation checking unit 7 retrieves the search result D50 (... word processor [name] / create document [name] / create [service] ..., ... / spec [name] / create [service name] from the search unit 4] … 、
… Document [name] Create [service name]…, Document [name] / Table [name] / Create [service]…) Keyword D1
While extracting the relationship between words from the above, the above search result D50
The relationship between the words is extracted for each of the documents 1 to 3, and the relationship is checked to see if this relationship matches the relationship of the keyword D1.

Also in this example, like the first invention, each analysis information forming the above retrieval result D50 for the keyword D1 "create a document" (document ← [category] ← create) Of the documents (document ←
[Case] ← Create) match, extracted from document (Specification ← [Case] ← Create) match, extract from document (Document ← [Case] ← Create) match, extract from document (Document ← [is case] ← created) does not match, and the check result D7 corresponding to these judgments is transmitted from the relationship check unit 7 to the priority calculation unit 9. In the column b of FIG. 4, this state is represented by a document in which a relationship match is recognized, a document, and a mode in which a circle is associated with the document.

The related word checking unit 8 sends the search result D50 (... word processor [first name] / document [first name] /
Create [Service] ..., Specification of [Name] / Create [Service]
..., ... Document [name] Create [service name] ..., Document [name] / Table [name] / Create [service] ...), and these are associated with the above keyword D1 in each document. The number and position (title, body) of words are checked (in this example, only the number of related words is output for simplicity), and the check result D8 is transmitted to the priority calculation unit 9.

Also in this example, as in the first invention, the related word "word processor" of the keyword D1 "create a document" is the analysis information (1 ... word processor [name]) about the document.
./Document [name] / create [service] ...), the check result D8 indicating that there is only one related word in the document is transmitted to the priority calculation unit 9. In the column b of FIG. 4, this state is shown in a form in which a number “1” indicating the number of related words is associated with the document in which the related word exists.

The priority calculation section 9 determines the search result D50 (
… In word processor [name] / create document [name] / create [support] ...,
… Specification [name] / Create [service name]…,… Document [name]
Create [service name] ..., Document [name] / Table [name] / Create [service] ...) and the above thesaurus expansion check result D6
(One document hit), the above relation check result D7
(Documents, documents, documents match), and when the related word check result D8 (the document includes one related word) is obtained, the search result D50 is formed by referring to these check results D6, D7, and D8. ~ Calculate each priority.

The weighting process for each check item at the time of calculating the priority is performed in the same manner as in the first invention. As a result, also in the present invention, in accordance with the contents shown in Table 1 (see column c of FIG. 4), as the priority information D9 of the search result D50, ([2100] ... word processor [name] / Create document [name] / Create [support] ..., [1500]
… Specification [Name] / Create [Service Name]…, [2000]… Document [Name] Create [Service Name]…, [0]… Document [Name] / Table [Name] / Create [Service] ] ...) is obtained, and this is transmitted from the priority calculation unit 9 to the display unit 10.

The display unit 10 displays the priority information D9 ([2100] ... word processor [name] / document [name]] of the search result D50.
Create / [Create] [...], [1500] ... Specification [Name] / Create [Support] ..., [2000] ... Create Document [Name] [Support],
[0]… In document [first name] / table [first name] / create [service]…)
Then, the text information storage unit 5 is searched for the text information corresponding to each of the search results D50, and this text information is then searched for the search result D corresponding to the text information.
The search result D10 is displayed to the user in accordance with the priority given to the item 50.

Specifically, as shown in the column d of FIG. 4, the points in the brackets [] in the priority information D9 are referred to, in order of increasing score, 1 ... Create a document with a word processor ... 2 ... Create document ... 3 ... Create specification ... 4 ... Create table with document ...

In the above embodiment, the case where the display unit 10 actually searches for text information and displays it in order of priority has been described. In addition to this, text information may be synthesized from the analysis information of the search result D50. However, in this case, it is not necessary to provide the text information storage unit 5.

As described above, according to the second aspect of the present invention, a search is performed for a target in which analysis information obtained by analyzing text information in advance is stored, and a priority is calculated. In the first invention, the text information is directly searched,
In order to calculate the priority, it was necessary to analyze the text information of the search result in the subsequent relation check and related word check. On the other hand, in the second invention, since the priority is calculated by using the analysis information which has been analyzed in advance in the text information, it is necessary to analyze the text information of the search result in the relation check or the related word check. There is no problem, and the processing can be performed at high speed.

[0074]

As described above, according to the first aspect of the present invention, the prioritization of information retrieval results by keywords is expanded so that it can be applied even when a keyword is a sentence or a clause, and Regarding the relationship with the corresponding information retrieval result, the check result is weighted by score calculation for each item of the semantic distance with the keyword, the match / mismatch of the relationship between the words with the keyword, and the description amount of the keyword. The search results are displayed according to the priority of each search result. Therefore, the search results can be prioritized with high accuracy by the comprehensive evaluation based on the above multiple item checks. You will be able to search for important information more efficiently by urging you to check from the highest.

In addition, in the second invention, instead of the method of the first invention, in which the text information is directly searched based on the keyword and then the above-mentioned item check is performed through analysis of the text information of the search result, Prepare analysis information obtained by analyzing text information in advance, first search this analysis information with a keyword, and then perform each item check described above to check the relation check and the relation check among the above item checks for the search result. Since the text information of the search result does not need to be analyzed in the related word check, a faster search operation can be realized.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of an information search device according to an embodiment of the first invention.

FIG. 2 is a conceptual diagram showing a flow of information search processing of the apparatus in FIG. 1 together with input and output examples for each component.

FIG. 3 is a schematic configuration diagram of an information search device according to an embodiment of the second invention.

FIG. 4 is a conceptual diagram showing the flow of information search processing of the apparatus in FIG. 3 with input and output examples for each component together.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Input part, 2 ... Thesaurus expansion part, 3 ... Thesaurus dictionary, 4 ... Search part, 5 ... Text information storage part, 6 ... Thesaurus expansion check part, 7 ... Relationship check part, 8 ... Related word check part, 9 ... Priority calculation section, 10 ... Display section, 11 ...
Analysis information storage

Claims (2)

[Claims] 1. A text information storage means for storing text information, an input means for inputting a search key, and a synonym obtained by expanding a synonym of the word of the search key from the search key input from the input means. Retrieval expression creating means for creating a retrieval expression including, related word storage means for storing related words, related word acquisition means for acquiring related words of the search key from the related word storage means, and the text information storage means And a synonym expansion checking unit for checking how many synonym expansion words in the search formula are included, according to the search result of the search unit. The relationship between words in the search results is
Relationship checking means for checking whether or not the relationship between words in the search key matches, and a relationship for checking the number and position of the search key and its related words included in each search result for each search result. A word check means, the synonym expansion check means, the relationship check means,
And a display unit for displaying each search result in accordance with the priority corresponding to each search result based on each check result of the related word checking unit. An information retrieval device characterized by: 2. A text information storage means for storing text information, an analysis information storage means for storing analysis information obtained by previously analyzing the text information, an input means for inputting a search key, and an input from the input means. From the search key, a search expression creating means for creating a search expression including a synonym obtained by expanding the word of the search key, a related word storage means for storing a related word, and the search from the related word storage means. A related word acquisition unit that acquires a related word of the key; a search unit that searches the analysis information storage unit by the search expression; and a number of words that are synonym-expanded in the search expression for each search result of the search unit. And a synonym expansion check means for checking whether or not there is a relation between words in each search result,
Relationship checking means for checking whether or not the relationship between words in the search key matches, and a relationship for checking the number and position of the search key and its related words included in each search result for each search result. A word check means, the synonym expansion check means, the relationship check means,
And a display unit for displaying each search result in accordance with the priority corresponding to each search result based on each check result of the related word checking unit. An information retrieval device characterized by: