JP2002123546A - Apparatus and method for document retrieval and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform relatively fast document retrieval. SOLUTION: This document retrieval apparatus which retrieves a document related to a given character string from plural documents is equipped with a means storing 1st word feature information which is generated for each document and shows information related to independent words included in the documents, an extracting means extracting independent words included in the character string, a means generating 2nd word feature information which shows information related to the independent words extracted by the extracting means, a means retrieving a document including all the independent words extracted by the extracting means, and a means sequencing retrieved documents according to the 1st word feature information on the independent word corresponding to the independent word extracted by the extracting means and the 2nd word feature information.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a document search technique.

[0002]

2. Description of the Related Art Various methods have been proposed for searching documents managed in a document database. FIG.
9 is a flowchart illustrating an example of a conventional document search method.

In step S201, a search condition is set, and in step S202, the search condition is analyzed to obtain a relationship between the words. For example, when the character string "rich experience" is set as the search condition in step S201, the words "rich" and "experience" are extracted in step S202, and the relationship between these words is acquired. .

In step S203, a document to be searched is analyzed in the same manner as in the query performed in step S202. As shown in FIG. 10, all managed documents 1
To n are search target documents, and document analysis and matching with search conditions are sequentially performed on each document.

As an example of document analysis, assuming that, for example, documents A to D shown in FIG. 11 are documents to be searched, the document A states that "Company A is affiliated with a company having abundant experience." And, the sentence "This partnership is a stepping stone to advance into the information-related field" is analyzed as shown in FIG.

In step S204, the ordering is performed by comparing the relationship between the document to be searched and the character string set as the search condition. Here, the ordering result is document A
> Document B> Document C> Document D, and as a search result, it is possible to present documents in order from the document having the highest relevance to the search condition.

As another method of the conventional document search method, there is a method as shown in a flowchart of FIG.

In step S301, search conditions are set. Here, it is assumed that the character string “rich experience” is set. In step S302, the search condition is analyzed. Here, the relationship between the words "rich" and "experience" and the keywords "rich" and "experience"
And extract

In step S303, step S302
A full-text search process is executed based on the keywords extracted in step (1), and documents containing both "abundance" and "experience" are searched.

In step S304, a full-text search result is obtained. In step S305, the search result document is analyzed, and the structure of the sentence is analyzed. Step S306
Then, the search condition structure and the sentence structure are compared, and a priority is added to the searched documents.

In step S307, step S305
Are executed for all the documents obtained as a result of the full-text search.

As a result, for example, the document A illustrated in FIG.
The order of documents D is document A> document B> document C, and document D is excluded from the search results. Thereafter, each document is presented in the order of the above as a search result.

As described above, some of the conventional search techniques can obtain a search result reflecting search conditions more than a simple full-text search.

[0014]

However, in general, the processing for analyzing a sentence structure is slow, so that the processing time for many texts such as a document becomes extremely long, and it has been difficult to put it to practical use. Therefore, it is more difficult to implement sentence structure analysis in a document search system that performs sentence structure analysis on many documents.

Therefore, an object of the present invention is to provide a document search apparatus, a document search method, and a document search method capable of performing a relatively high-speed document search.
It is to provide a recording medium.

[0016]

According to the present invention, there is provided a document retrieval apparatus for retrieving a document related to a given character string from a plurality of documents. A first indicating information on independent words contained in the document;
Means for storing word feature information, extraction means for extracting an independent word included in the character string, and means for creating second word feature information indicating information about the independent word extracted by the extraction means. Means for searching for a document including all of the independent words extracted by the extracting means from among the documents, and for the independent words extracted by the extracting means, among the independent words included in the searched document, A document search device, comprising: means for ranking the searched documents based on the first word feature information of the corresponding independent word and the second word feature information. Is provided.

Further, according to the present invention, there is provided a document search method for searching a document related to a given character string from a plurality of documents, wherein each of the documents relates to an independent word included in the document. Creating first word feature information indicating information in advance, storing the first word feature information in a recording medium, extracting an independent word included in the character string, and extracting information about the independent word extracted in the extracting step. Creating the second word feature information to be shown, searching the document for a document that includes all of the independent words extracted in the extracting step, and selecting the independent words included in the searched document. ,
The first word feature information of the independent word corresponding to the independent word extracted by the extraction means is read from the recording medium, and based on the first word feature information and the second word feature information. And ranking the retrieved documents.

Further, according to the present invention, in order to search for a document related to a given character string from a plurality of documents, a computer is included in the document created in advance for each document. Means for storing first word feature information indicating information related to an independent word; extracting means for extracting independent words included in the character string; second word feature indicating information related to the independent word extracted by the extracting means Means for creating information, means for searching for a document that includes all of the independent words extracted by the extracting means from among the documents, and among the independent words included in the searched documents, extracted by the extracting means A program functioning as means for ranking the searched documents based on the first word feature information of the independent word corresponding to the independent word and the second word feature information. Recording the recording medium is provided.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of a document search apparatus according to one embodiment of the present invention. The device includes an input device 1, a CPU 2, an output device 3, a storage device 4,
It is realized on a general-purpose computer provided with. The input device 1 includes, for example, a keyboard, and is used to input search conditions. The CPU 2 executes a later-described document search process according to the processing program 41 stored in the storage device 4. The output device 3 is a display or the like, and displays search results and the like. The storage device 4 is, for example, a hard disk, a RAM, or a combination thereof, and stores, in addition to the processing program 41 executed by the CPU 2, the word feature data 42, a document to be searched, and the like.

The document search apparatus of the present invention can be realized not only on a single computer as shown in FIG. 1, but also on a local network environment shown in FIG. 2 or an Internet environment shown in FIG. In this case, each client computer sets search conditions, displays search results, and the like, and the server performs search processing.

Next, the document search processing will be described. FIG. 4 is a flowchart showing the document search process.

In step S101, an input of a search condition is received from the input device 1. The search conditions are input by inputting a plurality of keywords (including logical expressions) or character strings such as natural sentences. Step S10
In step 2, an independent word is extracted from the search condition to create a logical expression for executing a full-text search.

For example, if the character string input as a search condition is “rich experience”, “rich” and “experience”
The word "rich" / "experience" is a conditional expression for executing a full-text search.

In step S103, step S102
A full-text search process is executed using the conditional expression created in step (1), and all documents including both "rich" and "experience" in the document are searched. The full-text search method is not limited. In the present embodiment, five documents A shown in FIG.
To E are assumed. Then, since the document D does not include “rich” and “experience”, it is excluded from the search results. Documents A to C and Document E are:
Since it includes "rich" and "experience", it is included in the search results.

Steps S104 to S106
Is a loop for ranking the searched documents, and the processes of steps S105 to S106 are executed for each searched document.

Here, in the document search device of this embodiment,
For each document to be searched, word characteristic data relating to an independent word included in the document is stored in the storage device 4. This word feature data is created in advance before searching, such as when a document is registered, and stored in the storage device 4.

The word feature data has, as shown in FIG. 6, information on independent words included in each document in a table format. Each independent word is classified by sentence (sentence number). Further, each independent word is attached in a table format with positional information in the sentence of the independent word, information on the use of the independent word, and information on auxiliary words such as particles attached to the independent word. FIG. 7 is a diagram illustrating information of some independent words included in the document A, and FIG. 8 is a diagram illustrating information of some independent words included in the document B.

In step S105, the above-mentioned word feature data is created for the character string given as the search condition, and the word corresponding to the independent word included in the search condition among the independent words included in the searched document. The word feature data on the independent word is acquired from the storage device 4.

In step S106, the degree of coincidence between the independent word in the search condition and the independent word in the searched document is calculated.
The ranking of each searched document is assigned.

In the present embodiment, ranking is performed in the following relationship.

Matching features of constituent words> Appear in same sentence
> Appearing in the same document For this reason, in the word feature matching degree calculation processing in step S106, the acquisition of the word feature data is first performed only for a sentence in which a word appears in the same sentence.

Therefore, of the documents A to E, the document D is excluded during the full-text search as described above, and the document E is given the lowest ranking from the viewpoint of the feature of the independent word.

The degree of coincidence can be calculated from the following viewpoints. -Distance score Based on the position information of the word feature data, the distance difference between the independent word in the search condition and the independent word in the document is normalized by the size of the sentence. The maximum is 1.

Distance score (SL) = 1- (distance between words / sentence size) Inflection type score Based on the utilization information of the word feature data, the independent words in the search condition and the independent words in the document are used. Calculate whether the shapes match. If they match, the score is 1. If the usage type is different, there is a cost associated with changing the usage type. -Attached words Based on the attached word information of the word feature data, calculate whether the attached words match. The attribute of the adjunct and the modifier type of the adjunct exist, and if they match, the score is 1
become. If different, there is a cost associated with the change.

Here, the search condition "rich experience" is as follows: Involved word: rich, position: 1, inflected form: continuous, adjunct information: none Recipient word: experience, position: 3, inflected form: stem, Attached word information: Analyzed as None.

When compared with the word feature data of the document A (FIG. 7), the match is 1 for all features.

When compared with the word feature data of document B (FIG. 8), the use of the search condition is the same, but the matching degree is smaller than that of document A because the receiving side is different. As described above, the degree of coincidence is calculated for the full-text search results, and each document is ranked. In this example, the order of the documents is A
>B>C> E. In the above example, only the position of the word was involved in the determination of the priority. In S107, S10
The search results are output in descending order of the priority assigned in step 6. Thus, the search processing ends.

As described above, by preliminarily registering feature data representing the features of words contained in a document to be searched, a high-speed search can be performed without performing sentence extraction processing or sentence structure analysis. Become.

Further, by simply comparing the feature data focused on the word and the feature data obtained from the conditional sentence given as the search condition, the degree of coincidence is calculated from the feature of the relationship between the words. Therefore, it is possible to output a search result with high accuracy according to the user's search intention at high speed. Further, since the search is based on the full-text search, search results with less omission can be obtained as compared with the case where the structural analysis is performed.

Next, an example relating to paraphrasing of an accessory word will be described.

Assuming that the search condition is a character string of "book about mountains", the analysis extracts an associated word "(related / union)" attached to the independent word "yama". On the other hand, for the following sentence in the document, the following is set as the feature data of the attached word part attached to the word “mountain”. 1) Books about mountains (related / continuous) 2) Books written about mountains (related / continuous) 3) Books related to mountains (related / continuous) 4) Books related to mountains (location / continuous) 5) Yamano-no book (-, continuous) 6) Yama-he book to (place, continuous) For the above sentences (1) and (3), the method of expression is different, but the attributes of the attached words and the modifier types are different. It can be determined that they are the same and almost synonymous. In the sentence (2),
Although there is an attribute, the subject of the decoration may be wrong, and the matching degree decreases. Next, the sentence of (4) differs in the attribute of the attached word, and the degree of coincidence further decreases. In the sentence (5), it can be seen that the attributes of the attached words are different, but the objects of the modification are the same. In (6), the objects to be modified are the same, but the attributes of the attached words are different.

As described above, (1), (3)> (2)
> (4), (5)> (6).

An object of the present invention is to supply a storage medium (or a recording medium) on which a program code of software for realizing the functions of the above-described embodiments is recorded to a system or an apparatus, and to provide a computer (or a computer) of the system or the apparatus. It is needless to say that the present invention can also be achieved by a CPU or an MPU) reading and executing the program code stored in the storage medium. In this case, the program code itself read from the storage medium implements the functions of the above-described embodiment, and the storage medium storing the program code constitutes the present invention. Also,
By executing the program code read by the computer, not only the functions of the above-described embodiments are realized, but also an operating system (OS) running on the computer based on the instructions of the program code.
It goes without saying that a case where the functions of the above-described embodiments are implemented by performing some or all of the actual processing, and the processing performs the functions of the above-described embodiments.

Further, after the program code read from the storage medium is written into the memory provided in the function expansion card inserted into the computer or the function expansion unit connected to the computer, the program code is read based on the instruction of the program code. Needless to say, the CPU included in the function expansion card or the function expansion unit performs part or all of the actual processing, and the processing realizes the functions of the above-described embodiments.

As described above, according to the present invention,
A relatively high-speed document search can be performed.

[Brief description of the drawings]

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

FIG. 2 is a diagram showing another configuration example in which the document search device of the present invention is realized.

FIG. 3 is a diagram showing another configuration example in which the document search device of the present invention is realized.

FIG. 4 is a flowchart illustrating a document search process according to an embodiment of the present invention.

FIG. 5 is a diagram illustrating an example of a document to be searched;

FIG. 6 is a diagram showing word feature data.

FIG. 7 is a diagram showing word feature data.

FIG. 8 is a diagram showing word feature data.

FIG. 9 is a flowchart illustrating an example of a conventional document search method.

FIG. 10 is a diagram showing a process for each document to be searched in the conventional document search method.

FIG. 11 is a diagram illustrating an example of a search target.

FIG. 12 is a diagram showing an analysis result of document A in FIG. 11;

FIG. 13 is a flowchart showing another example of a conventional document search method.

Claims (7)

[Claims] 1. A document retrieval apparatus for retrieving a document related to a given character string from a plurality of documents, comprising: information on an independent word included in the document, which is created in advance for each document. Means for storing the first word feature information shown, extraction means for extracting independent words included in the character string, and creation of second word feature information indicating information about the independent words extracted by the extraction means Means for searching for a document including all of the independent words extracted by the extracting means from the documents; and extracting the independent words included in the searched words from the independent words included in the searched document by the extracting means. Means for ranking the retrieved documents based on the first word feature information of the independent word corresponding to the independent word and the second word feature information. Document search device. 2. The document search device according to claim 1, wherein both the first word feature information and the second word feature information are syntactic information of the independent word. 3. The first word feature information includes information indicating the position of the independent word in the document, and the second word feature information includes information indicating the position of the independent word in the search character string. 2. The document search device according to claim 1, wherein information indicating a position in the document is included. 4. The document search apparatus according to claim 1, wherein the first word feature information and the second word feature information each include information on an adjunct to the independent word. 5. The document search device according to claim 1, wherein the first word feature information and the second word feature information each include information on the use of the independent word. 6. A document retrieval method for retrieving a document related to a given character string from a plurality of documents, comprising: for each of the documents, a first information indicating information on an independent word included in the document. A step of creating word feature information in advance and storing the word feature information in a recording medium; an extraction step of extracting an independent word included in the character string; and a second word feature indicating information on the independent word extracted in the extraction step Creating information; and searching the document for a document that includes all of the independent words extracted in the extraction step; extracting the independent words included in the searched document by the extraction unit Reading out the first word feature information of the independent word corresponding to the independent word from the recording medium,
A step of ranking the searched documents based on the word feature information and the second word feature information. 7. A computer for retrieving a document related to a given character string from a plurality of documents, the computer showing information on an independent word included in the document, which is created in advance for each document. Means for storing first word feature information; extracting means for extracting independent words included in the character string; means for creating second word feature information indicating information about the independent words extracted by the extracting means; Means for searching a document including all of the independent words extracted by the extracting means from the document, corresponding to the independent words extracted by the extracting means among the independent words included in the searched document A recording medium storing a program functioning as means for ranking search documents based on the first word feature information of the independent word and the second word feature information.