JPH09198395A - Document retrieval device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To execute efficient retrieval through the use of a keyword from the next time even if the inputted keyword is not registered in an index. SOLUTION: When a retrieval expression is inputted from an input part 1, a keyword extraction part 2 refers to a morpheme analysis dictionary 3 based on the retrieval expression inputted from the input part 1, analyzes a morpheme, extracts the keyword and inputs the keyword to a retrieval part 4. The retrieval part 4 retrieves the index by setting an index storage part 5 to be an object with the keyword as a retrieval key. When the index retrieval result cannot be obtained, a whole sentence retrieval part 8 retrieves the whole sentence of the document in a document storage part 6 by means of searching a character string for the retrieval expression. When the whole sentence retrieval result can be obtained, an addition/registration part 9 registers the word of the retrieval expression in the morpheme analysis dictionary 3 and the index 5 as one word.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document search in which an index is created for a document to be registered and a document is specified by searching the inputted keyword using the index when searching. Regarding the device.

[0002]

2. Description of the Related Art As a typical search method for searching a document, a full-text search method for searching a full-text using a character string search and an index for a document to be registered are prepared. An index search method is known in which a document is specified by performing a search for an input keyword using the index.

The full-text search method based on the character string search requires a considerably longer search time than the index method in which a keyword is added in advance. Further, if the collation of the search condition logical expression is specified in the case where a character string search such as logical product or logical sum of the character strings A and B is specified, the speed further decreases. Moreover, this full-text search method requires a long search time each time even if the same search expression is given.

On the other hand, in the index search method, a specific keyword is extracted in advance from the document to be registered by the keyword extracting means and registered in the index, and at the time of search, the keyword extracting means is used from the input search expression. Basically, the search key is extracted and the document is specified by referring to the index corresponding to this search key. Therefore, it is sufficient to search only the document indicated by the index, which enables fast and flexible search. It will be possible.

However, in the index search method, the advantage of not having to search the entire text is utilized only when the keyword is successfully extracted from the search expression and the keyword is registered in the index. It goes without saying that it will be done.

However, in document retrieval, for example,
When the search expression is a compound word, a situation may occur in which the keyword is not successfully extracted. In order to eliminate such inconvenience, conventionally, in JP-A-3-116375,
When the index is registered, the compound word is divided into simple words and registered, and at the time of search, by applying a predetermined search condition between each word of the keywords composed of the compound words, the hit ratio of the input keyword to the index can be determined. An information retrieval device designed to increase the number has been proposed.

In this conventional apparatus, for example, in the case where the keyword "noise prevention regulations" is added,
"Noise", "Prevention",
In addition to the search by each keyword "regulation", the search by the keyword "noise regulation" can be realized.

In this conventional apparatus, by dividing all simple words into keywords and registering them in the index, even if a compound word comes in, it contributes to the improvement of the point that the input keyword is not well extracted. However, we could not deal with the unsearchability caused by the keyword not being registered in the index.

Therefore, as a measure for eliminating such inconvenience, a method of registering the keyword to be extracted in a user dictionary or a specialized dictionary in advance can be considered. However, even with this method, there is no difference from the above method in that the search cannot be performed if the desired keyword is not included in the user dictionary or the specialized dictionary. Moreover, in this method, since a required keyword is registered in the dictionary each time a new document is registered, the keyword registered later is not reflected in the already registered database, Despite registering with great care, I could not search using this keyword.

[0010]

As described above, the conventional document retrieval apparatus employs the index retrieval method, so that the same retrieval expression is used every time in full-text retrieval and the retrieval is performed for a long time. Although it can reduce waste, on the other hand, due to the peculiar circumstances of the index search method, even if the same keyword as the search expression exists in the document, if that keyword is not registered in the index, the document cannot be searched. There was a problem.

The present invention has been made in view of the above circumstances, and from the next time onward, even if the input keyword is not registered in the index, a document search device capable of performing an efficient search using the keyword is provided. The purpose is to provide.

[0012]

According to the present invention, there are provided a document storage means for storing a document, an index storage means for storing an index of a document stored in the document storage means, and the index storage means based on a search key. In a document search device, which is provided with an index search unit that searches and identifies a document corresponding to the search key, and performs document search based on an index search result of the index search unit with respect to the search key, the document using the search key. A text search unit that searches the document stored in the storage unit and specifies a document including the search key; and if the index search unit does not specify the document corresponding to the search key, the text search unit A document stored in the document storage means is searched, and the document is specified by the document search. If, the search key is characterized in that it comprises the additional registration means for additionally registering the index storage means in association with the document information of the document specified by the retrieval key.

Preferably, the present invention extracts a search keyword by morphologically analyzing a morpheme analysis dictionary storing morpheme analysis information and a search expression input at the time of search using the morpheme analysis information, and the search keyword And a keyword extracting means for providing the index search means as the search key, wherein the additional registration means, when additionally registering the search key in the index storage means, stores the search key and the part-of-speech information for the search key. It is characterized in that it is additionally registered in the morphological analysis dictionary.

[0014]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. FIG. 1 is a block diagram showing a schematic configuration of a document search device according to the first embodiment of the present invention. This document search device includes an input unit 1, a keyword extraction unit 2, a morphological analysis dictionary 3, a search unit 4, an index storage unit 5, a display unit 6, a document storage unit 7,
A full-text search unit 8 and an additional registration unit 9 are provided.

The morphological analysis dictionary 3 stores, for example, morphological analysis information in which headwords are associated with part-of-speech information, as shown in FIG. This morphological analysis information is used for morphological analysis of the search expression input at the time of search.
In the index storage unit 5, for example, as shown in FIG.
An index formed by associating the keyword extracted at the time of document registration in the document storage unit 7 with the document information (document ID) of the document including the keyword is stored. The document storage unit 7 stores a document including each keyword in the index.

In the document retrieval apparatus of the present invention, as a method for retrieving a document in the document storage unit 7, a method of retrieving based on an index stored in the index storage unit 5 and a document storage unit regardless of the index are used. There is a method of searching the entire text of the documents in 7 by a character string search.
The search unit 4 has a search function by the former search method, and the full-text search unit 8 has a search function by the latter method.

Next, the document search processing in this document search device will be described with reference to the flowchart shown in FIG. First, when a document search is started, a user selects a search formula in consideration of the content of the document to be searched, and the search formula is input from the input unit 1 (step 1
01). This search formula is input from the input unit 1 to the keyword extraction unit 2
Is forwarded to The keyword extracting unit 2 refers to the morpheme analysis dictionary 3 based on the above-mentioned search expression, performs morpheme analysis, and extracts words such as nouns constituting the search expression as keywords. A word as a keyword extracted by this morphological analysis is input from the keyword extracting unit 2 to the searching unit 4.

The search section 4 performs an index search on the registered information in the index storage section 5 using the above-mentioned word input from the keyword extraction section 2 as a search key (step 102), and searches for the index search target. It is determined whether the search key has been extracted (step 10).
3).

If a search key for index search is extracted (YES in step 103), the search unit 4 then reads out the indexes corresponding to all the extracted search keys, and the document information in all the indexes. An AND operation between them is performed (step 104). Then, the search unit 4 displays the index search result on the display unit 6 according to the calculation result of the logical product calculation (step 105), and ends the series of search processing.

On the other hand, when the search key for index search is not extracted after the index search is started (NO in step 103), then the search unit 4
In step 101, all the character strings of the search expression received from the keyword extracting unit 2 are transferred to the full-text searching unit 8. here,
The full-text search unit 8 uses the entire character string of the input search expression as one word, and with respect to this search expression, the document storage unit 7
While performing a full-text search for the document registered in (step 110), it is determined whether or not the search result of the full-text search is obtained (step 111). Here, when the full-text search result is not obtained, that is, when the search result is “0” (step 111 NO), the search process is immediately terminated.

On the other hand, when the above-mentioned full-text search result is "1" or more (YES in step 111), the full-text search unit 8 notifies the additional-registration unit 9 of the full-text search result. The additional registration unit 9 registers the character string of the search expression used in the full-text search as one word and the part-of-speech information as a noun in the morphological analysis dictionary 3 based on the notified search result, and the index storage unit 5 For, the character string of the above search formula in the keyword field, and the document ID searched by the search formula
Are registered in the document information fields (step 112).
Next, the full-text search unit 8 displays the full-text search result on the display unit 7.
Is displayed (step 105), and a series of search processing is completed.

Next, regarding a more specific processing operation of the document search, for example, a case where a character string "application of natural language processing technology" is input as a search expression and "DP30
The case where the product name "0" is input will be described in detail.
In this case, the content of the registration information of the morphological analysis dictionary 3 is as shown in FIG.
It is assumed that the content of the registration information is as shown in FIG.

First, when a search expression "application of natural language processing technology" is input from the input unit 1, the keyword extracting unit 2 performs morphological analysis on this search expression by referring to the morphological analysis dictionary 3, "Nature / language / processing / technology / of /
After obtaining the analysis result of "application", the noun "natural",
“Language”, “processing”, “technology”, and “application” are extracted as keywords and transferred to the search unit 4.

Next, the search unit 4 is connected to the keyword extraction unit 2
An AND search of the index in the index storage unit 5 is performed using each of the keywords received from the above. At this time, based on the registered contents of the index storage unit 5 (see FIG. 4), first, as document information of a document including each of the above keywords, “1, 3, 5”, “1, 7, 8”, “1, 5, 9”. ],
By obtaining "1, 6, 8" and "1, 5, 6", and then performing an AND operation on each of these document information, an AND search result of document information "1" is obtained.

At this time, the display section 6 displays the index search result in the form of search result: document information "1", for example. The display method of the search result can be modified in various ways such as reading the document corresponding to the document information “1” which is the AND search result from the document storage unit 7 and displaying the content. .

On the other hand, when the search expression "DP300" is input from the input unit 1, the keyword extracting unit 2 refers to the morphological analysis dictionary 3 to perform morphological analysis. , The keyword “DP300” is not registered (see FIG. 3), the search formula “D
When inputting "P300", the keyword extraction result is not returned from the keyword extraction unit 2 to the search unit 4.

Therefore, the search unit 4 cannot perform the index search itself for the input keyword, and the keyword search result at this time is "0". The retrieval unit 4 which has obtained the retrieval result “0” fetches the retrieval formula “DP300” from the keyword extraction unit 2 and transfers it as a word to the full-text retrieval unit 8 as it is.

The full-text search section 8 uses the search expression "DP300" provided from the search section 4 as a character string for searching, and performs a full-text search by a character string search for all documents registered in the document storage section 7. To do.

Here, the document 2 registered in the document storage unit 7
And the document 4 include the character string of the search expression "DP300", and the full text search result is document information "2, 4".
Shall be obtained. FIG. 5 shows the document contents of Document 2 among them, and the character string “DP300” is recognized in the second line of the document.

The result of the full-text search is displayed on the display unit 6 in the form of search result: document information "2, 4", for example. When displaying the search result in this case, in addition to the direct display of the document ID,
A method of reading the document having the document ID from the document storage unit 7 and displaying the content may be used.

As a result of the above-mentioned full-text search, the document information "2
4 ”is obtained, and the additional registration unit 9 that recognizes that the full-text search result is not“ 0 ”prepares a search environment using the above-described search expression“ DP300 ”based on the search result. The information is registered in the morphological analysis dictionary 3 and the index storage unit 5.

Specifically, for the morphological analysis dictionary 3, the additional registration unit 9 describes "DP300", which is the character string used for the full-text search at this time, in the entry word column, and the entry word The information "noun" is described in the column of part-of-speech information corresponding to the column. For the index storage unit 5, the character string "DP300" is described in the keyword column, and "2, 4" is described in the document information column corresponding to this keyword column.

FIG. 6 shows an example of registration information of the morphological analysis dictionary 3 after the additional registration, and similarly,
FIG. 7 shows the index storage unit 5 after the additional registration.
7 shows an example of the registration information of.

Morphological analysis dictionary 3 as shown in these figures
Under the registration environment of the index storage unit 5, even when "DP300" is input as a search expression thereafter, by morphological analysis using the morphological analysis dictionary 3,
The keyword "DP300" can be reliably extracted, and the index AN by this keyword "DP300"
As a result of the D search, a search result of document information “2, 4” is obtained, and a short-time search that does not depend on the full-text search can be realized.

That is, when the search expression "DP300" is input from the input unit 1 after the above additional registration (step 101), the keyword extracting unit 2 causes the morphological analysis dictionary 3 to operate.
To start the morphological analysis. In this morphological analysis, "DP300" as a noun is extracted as a keyword based on the registered contents of the morphological analysis dictionary 3 shown in FIG.

This keyword "DP300" is passed to the search unit 4 and is attached to the index search (step 102). Here, the search unit 4 is based on the registered contents as shown in FIG. 7 in the index storage unit 5 at this time.
The keyword "DP300" is found (YES at step 103), and the AND search of the document information corresponding to this keyword "DP300" is performed (step 104).

At this time, the AND search result of the document information "2, 4" for the keyword "DP300" is obtained from the registered contents of the index storage unit 5 (see FIG. 7). As a result, the document information "2, 4" is displayed on the display unit 6 as the index AND search result for the keyword "DP300" (step 105). Further, at this time, when a predetermined search instruction is given from the user, the search unit 4 searches the document “2” or “4” or both documents from the document storage unit 7, and displays the document contents. Display in 6.

As described above, according to the present invention, since the specified search expression is not registered in the morphological analysis dictionary 3, an appropriate keyword is not extracted, and when the index search result by the above search expression is not obtained, the above search is performed. When a full-text search is performed on the expression by a character string search and a search result is obtained by this full-text search, the word of the above-mentioned search expression is regarded as one word and added to the morphological analysis dictionary 3 and the index storage unit 5. It is the one to be registered.

In the normal index method, if the search key is not registered in the index, the search result is not returned even if the desired keyword is present in the document, but the above-mentioned full-text search function and additional registration function are provided. Further, according to the configuration of the present invention, once the entire text can be searched using the search key that is not registered in the index, the index search can be performed efficiently from the next time.

If the search key is not registered in the index and the search result is not returned even if the desired keyword is present in the document, a user dictionary or a specialized dictionary is prepared for the keyword. However, if the desired keyword is not included in the keyword, the same is true, and the flexible search cannot be performed unless the search key necessary for the search is registered.

With respect to this point, in the present invention, when extracting the words to be additionally registered, since all the documents registered in the document storage unit 7 before that are searched, the words after the additional registration are As a keyword, it can be surely reflected in the already registered document.

In the first embodiment, if no search result is obtained in the full-text search after a proper keyword is not extracted from the search expression (step 1).
No. 20)), the example in which the search operation is immediately terminated is disclosed, but in this case as well, the search expression is changed to the morphological analysis dictionary 3
You may register only for.

According to this processing, when this search expression is input next time, the keyword is extracted by the morphological analysis with reference to the morphological analysis dictionary 3 by the keyword extracting unit 2, and then the index search is performed based on this keyword. When I do, I fall into the situation that the index search result cannot be obtained. Here, the fact that the search expression is registered in the morphological analysis dictionary 3 and not registered in the index storage unit 5 means that the full-text search has already been performed for this keyword, and the result of the full-text search and the search result Means that was not obtained. Therefore, in this case, the search unit 4 determines the meaning from the search result and cancels the full-text search after step 110, which requires a long time, so that the processing efficiency can be improved.

In the first embodiment described above, when the index search result by the search formula is not obtained, the full text search is performed by the character string search for the search formula. The search is not necessarily limited to the full-text search, and the same effect can be expected even if a part of the document is searched. This also applies to other embodiments described later.

Next, another embodiment of the present invention will be described. In the first embodiment, the morphological analysis dictionary 3 and the index storage unit 5 are used as additional registration destinations of the search expression according to the full-text search result after an appropriate keyword is not extracted from the search expression. However, as a second embodiment, a configuration is possible in which a user dictionary is provided separately from the morphological analysis dictionary 3 and additionally registered. FIG.
FIG. 3 is a block diagram showing a basic configuration of the document search device according to the second embodiment, which has a configuration in which a user dictionary 10 is added to the device in FIG.

FIG. 9 is a flow chart showing the search operation of the document search device according to the second embodiment. In this document search device, a full-text search is performed by the full-text search unit 8 using the entire character string of the input search expression as one word (step 110), and when the result of this full-text search is "1" or more (step 110). 111 YES), the additional registration unit 9 performs processing of registering the character string of the search expression, which is a keyword used in the full-text search, as one word and the part-of-speech information as a noun in the user dictionary 10, and at the same time, the index storage unit 5 , The character string of the above search formula is used in the keyword field, and the document I searched by this search formula is used.
D is registered in each of the document information fields (step 12)
0).

According to this document retrieval apparatus, by having a dedicated storage unit (user dictionary 10) as an additional registration destination, it is not necessary to directly additionally register unnecessary information in the morphological analysis dictionary 3, and a new information is newly added. When it is desired to create other data from the original dictionary data, the user dictionary 10 can be turned off to perform registration under the first environment.

Further, as a third embodiment of the present invention,
It is conceivable that the search expression has a corresponding function when it is given by a logical operation expression of AND (logical product). For example, if a search expression is given as "A &B", it is possible that one of the search expressions is indexed and the other is not.
In the device according to the third embodiment, even in the above case, the full-text search is performed by the full-text search unit 8 for one search expression that is not registered in the index, and the above-mentioned search is performed according to the search result. The other of the expressions is added to the morphological analysis dictionary 3 (or the user dictionary 1 by the additional registration unit 9).
0) and a function of additionally registering in the index storage unit 5 are added. In this way, by adopting a configuration that can also be applied to logical operations, the use of search is expanded and more efficient search can be realized.

Further, in the first embodiment, the part of speech of the word newly registered through the full text search is used as the noun, but in the fourth embodiment of the present invention, the morphological analysis in the morphological analysis dictionary 3 is performed. The user may specify from a plurality of parts of speech that can be used for. In this case, it may be possible to adapt to use a default setting (for example, a noun) when an appropriate part-of-speech is not known. According to the apparatus according to the fourth embodiment, a user who wants to classify parts of speech to some extent can specify from the prepared parts of speech, which improves accuracy in morphological analysis of a newly registered word. , The keywords can be extracted accurately.

[0050]

As described above, according to the present invention,
When no index search result is obtained when performing an index search based on the search key, the search is continued by switching to document search by character string search with the above search key as one word, and the result of this document search is obtained In addition, since the above search key is additionally registered in the index as one word together with the corresponding document information, when the same search key is given from the next time, the index already registered corresponding to the search key will be found. hand,
An efficient search can be realized by the index search method.

In particular, a morpheme analysis dictionary is provided, and a search keyword is extracted by performing a morpheme analysis on a search expression input at the time of search using the morpheme analysis dictionary, and the index search is performed using the search keyword as the search key. In the device adopting the method, when additionally registering the search key for the index, by additionally registering the search key and the part of speech information for the search key also in the morphological analysis dictionary, a keyword based on the search formula The extraction and the index search based on the keyword are ensured to contribute to the realization of a more efficient search.

[Brief description of drawings]

FIG. 1 is a diagram showing a schematic configuration of a document search device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing an outline of a search operation in the device shown in FIG.

FIG. 3 is a diagram showing an example of registered contents of a morphological analysis dictionary in the device shown in FIG.

FIG. 4 is a view showing an example of index registration contents in the device in FIG.

5 is a diagram showing an example of a full-text search result by the device in FIG.

6 is a diagram showing an example of registered contents of a morphological analysis dictionary after update registration based on a full-text search result in the apparatus in FIG.

FIG. 7 is a diagram showing an example of index registration contents after update registration based on a full-text search result in the apparatus in FIG.

FIG. 8 is a diagram showing a schematic configuration of a document search device according to another embodiment of the present invention.

9 is a flowchart showing an outline of a search operation in the apparatus shown in FIG.

[Explanation of symbols]

1 ... Input unit, 2 ... Keyword extraction unit, 3 ... Morphological analysis dictionary, 4 ... Search unit, 5 ... Index storage unit, 6 ... Display unit, 7 ... Document storage unit, 8 ... Full text search unit, 9 ... Additional registration unit 10 ... User dictionary

Claims (2)

[Claims] 1. A document storage means for storing a document, an index storage means for storing an index of the document stored in the document storage means, and a search for the index storage means based on a search key, which corresponds to the search key. And a document search device for performing a document search based on an index search result of the index search unit with respect to the search key, the document stored in the document storage unit using the search key. A text search means for searching and specifying a document including the search key; and a document stored in the document storage means by the text search means when a document corresponding to the search key is not specified by the index search means. When a document is identified by the document search, the search key is searched for. Document search apparatus characterized by comprising the additional registration means for additionally registering the index storage means in association with the document information of the document specified by the key. 2. A morphological analysis dictionary that stores morphological analysis information, and a morphological analysis of a search expression input at the time of search using the morphological analysis information to extract a search keyword, and the search keyword is used as the search key. A keyword extraction unit for giving the index search unit, wherein the additional registration unit adds the search key and the part-of-speech information for the search key to the morphological analysis dictionary when additionally registering the search key in the index storage unit. The document search device according to claim 1, wherein the document search device is additionally registered.