JP2000339323A - Method, device and system for retrieving document and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To scalably process mass document indexes and mass user queries. SOLUTION: This device has a user query processing part A-1 which performs morphological analysis of a user query and records it in a shared memory S-1, an object document processing part A-2 which performs morphological analysis of an inputted processing object document, prepares a document index and records it on a document index database D-3, registration query processing A-3 which reads a query content from the shared memory, retrieves it from the document index database and records proper document information on the shared memory when the document information fitting the user query is extracted and a retrieval result processing part A-4 which reads the document information being a retrieval result recorded on the shared memory and outputs it to the user.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document search method and a document search apparatus using a shared memory when searching for a document.
The present invention relates to a document search system and a computer-readable recording medium that stores a program for causing a computer to execute the document search method.

[0002]

2. Description of the Related Art Conventionally, a document search apparatus accumulates a document index in one apparatus and processes a search query using the document index. For this reason, when the document index becomes large or a large number of user queries occur, a state where processing cannot be performed occurs.

[0003]

SUMMARY OF THE INVENTION The present invention has been made in view of such circumstances, and a document search method and a document search method capable of scalably processing a large number of document indexes and a large number of user queries. It is an object of the present invention to provide a computer-readable recording medium that records a program for causing a computer to execute the apparatus, the document search system, and the document search method.

[0004]

In order to achieve the above object, according to the first aspect of the present invention, a user query is input,
The input user query is morphologically analyzed using a morphological analysis dictionary to determine words described in the user query, and the part of speech, inflected form, etc. of each word are determined, and the morphologically analyzed query is determined. User query processing to record in the shared memory space, a document to be processed is input, the input document is subjected to morphological analysis using a morphological analysis dictionary, and words of each sentence described in the document are determined. Along with the target document processing for recording the words and document names included in each sentence in a document index database as an index of each document, and constantly checking whether or not there is a query stored in the shared memory space, the shared memory At the same time as confirming that the query was recorded in the space, the contents of the query were read from the shared memory space, and a group of words matching the query was read. When the document index database is searched from the document index database and the document information matching the user query is extracted, the registration query processing for recording the document information in the shared memory space, and the registration query processing in the shared memory space It is constantly checked whether or not the document information recorded by the user is recorded. If appropriate document information is described as a search result with respect to the user query, the document is a search result recorded in the shared memory space. And a search result process for reading out information and outputting the information to a user.

According to the first aspect of the present invention, a user query is input, and the input user query is morphologically analyzed using a morphological analysis dictionary to determine a word described in the user query. User query processing for determining the part of speech, inflected form, etc. of each word, and recording the morphologically analyzed query in a shared memory space; inputting a document to be processed; and a morphological analysis dictionary for the input document. Morphological analysis using, to determine the word of each sentence described in the document, and the target document processing to record the words and document names included in each sentence as an index of each document in the document index database, It is constantly checked whether or not there is a query stored in the shared memory space, and at the same time it is confirmed that the query has been recorded in the shared memory space, When the content of the query is read from the memory space, a document having a word group that matches the query is searched from the document index database, and when document information matching the user query is extracted, the document information is stored in the shared memory space. In the registration query process of recording the document information, it is constantly checked whether or not the document information recorded by the registration query process is recorded in the shared memory space, and appropriate document information is described as a search result for the user query. In this case, there is a search result process for reading out the document information, which is a search result recorded in the shared memory space, and outputting it to the user, so that a large amount of document indexes and a large number of user queries can be processed scalably. can do.

According to a second aspect of the present invention, there is provided a document index database in which different document indexes are stored for each of a plurality of distributed document search devices, and the plurality of document search devices share a shared memory space. A document search method in a document search system that performs a document search by using a plurality of document search devices, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space. At the same time as confirming that the query has been recorded, the content of the query is read from the shared memory space, a document having a word group that matches the query is searched from the document index database, and document information matching the user query is obtained. When extracting the document information, the registration query processing unit that records the document information in the shared memory space; Whether the document information recorded by the registration query processing in the memory space is recorded constantly confirmed, as the search result for the user query,
When appropriate document information is described, the relevant document information which is a search result recorded in the shared memory space is read out,
And a search result process for outputting to the user.

According to the second aspect of the present invention, there is provided a document index database in which different document indexes are stored for each of a plurality of distributed document search apparatuses, and the plurality of document search apparatuses occupy a shared memory space. A document search method in a document search system for performing a shared document search, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space. At the same time as confirming that the query was recorded in the space, the contents of the query were read from the shared memory space, and a document having a word group matching the query was searched from the document index database, and the document was matched with the user query. When extracting the document information, a registration query processing unit that records the document information in the shared memory space, It is constantly checked whether or not the document information recorded by the registration query process is recorded in the shared memory space, and, as a search result for the user query, when appropriate document information is described, the shared memory space is And a search result process for reading out the document information that is a recorded search result and outputting the document information to the user. Can be fused. This is because the document indexes stored in the document index database of each search device are different, that is, the document indexes are created in a stripe shape.

According to a third aspect of the present invention, there is provided a document index database in which the same document index is stored for each of a plurality of distributed document search devices, and the plurality of document search devices share a shared memory space. A document search method for performing a document search by performing a document search, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space; At the same time as confirming that the query has been recorded, the content of the query is read from the shared memory space, a document having a word group that matches the query is searched from the document index database, and each of the other distributed documents is searched. When the search device records the search result in the shared memory space, it is confirmed that each of the other document search devices does not record the search result. Registration query processing to prevent the other document search devices from recording search results during recording, and whether or not document information recorded by the registration query processing is recorded in the shared memory space. Is constantly checked, and when appropriate document information is described as a search result for the user query, a search result process for reading out the document information, which is a search result recorded in the shared memory space, and outputting it to the user And the following.

According to the third aspect of the present invention, there is provided a document index database in which the same document index is stored for each of a plurality of distributed document search devices, and the plurality of document search devices are provided in a shared memory space. A document search method in a document search system for performing a document search by sharing documents, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space, and At the same time as confirming that the query was recorded in the memory space, the content of the query was read from the shared memory space, a document having a word group matching the query was searched from the document index database, and another distributed document was searched. When each document search device records a search result in the shared memory space, each of the other document search devices does not record the search result. Registration query processing to prevent each of the other document search devices from recording a search result during recording, and whether the document information recorded by the registration query processing is recorded in the shared memory space. Whether the document information is described as a search result for the user query and appropriate document information is described as a search result for the user query, the search result is output to the user by reading the document information, which is the search result recorded in the shared memory space. Since it has the result processing, it is possible to use the search result that responds at the highest speed among the respective document search devices constituting the document search system, and fuse the search results that respond at the highest speed. This is because the document indexes stored in the document index database of each search device are the same, that is, the document indexes are mirrored.

According to a fourth aspect of the present invention, a user query input unit for inputting a user query, and a user query input by the user query input unit are morphologically analyzed using a morphological analysis dictionary. And a query morphological analysis unit that determines the part of speech, inflected form, etc. of each word, and a query recording unit that records a user query morphologically analyzed by the query morphological analysis unit in a shared memory space. A user query processing unit having a, a target document input unit for inputting a document to be processed, and a morphological analysis of the document input by the target document input unit using a morphological analysis dictionary, which is described in the document. The document index that determines the words of each sentence and records the words and document names contained in each sentence as a document index in the document index database. A target document processing unit having a query creation unit, a query confirmation unit that constantly checks whether there is a query stored in the shared memory space, and the query confirmation unit confirms that the query has been recorded. At the same time, the content of the query is read out, and a document index search unit for searching a document having a word group matching the query from the document index database, and the document index search unit searches for the document information matching the user query. When a is extracted, a registration query processing unit having a search result recording unit that records the document information in the shared memory space,
A search result confirmation unit that constantly checks whether or not the document information recorded by the search result recording unit is recorded in the shared memory space; and A search result processing unit having a search result output unit that reads out the document information that is a search result recorded in the shared memory space when the document information is described, and outputs the document information to a user. I do.

According to the fourth aspect of the present invention, a user query input unit for inputting a user's query, and a user query input by the user query input unit are morphologically analyzed using a morphological analysis dictionary. A query morphological analysis unit that determines words described in the query and also determines the part of speech, inflected form, etc. of each word, and a query record that records a user query morphologically analyzed by the query morphological analysis unit in a shared memory space A user query processing unit having a unit, a target document input unit for inputting a document to be processed, and a morphological analysis of the document input by the target document input unit using a morphological analysis dictionary, and describing the document in the document. Sentence that determines the words of each sentence and records the words and document names included in each sentence as a document index in the document index database. A target document processing unit having an index creation unit, a query confirmation unit for constantly confirming whether or not there is a query stored in the shared memory space, and the query confirmation unit confirming that the query has been recorded. At the same time, the contents of the query are read, and a document index search unit that searches the document index database for a document having a word group that matches the query, and document information that is searched by the document index search unit and matches the user query Is extracted, the registered query processing unit having a search result recording unit that records the document information in the shared memory space, and whether the document information recorded by the search result recording unit is recorded in the shared memory space A search result confirming unit for constantly confirming whether or not the search result is obtained, the search result confirming unit as a search result for the user query; When appropriate document information is described, a search result processing unit having a search result output unit that reads out the document information, which is a search result recorded in the shared memory space, and outputs it to the user, Scalable processing of document indexes and large numbers of user queries.

According to a fifth aspect of the present invention, there is provided a document search system in which a plurality of distributed document search devices share a shared memory space to perform a document search. A query index unit that has a document index database in which different document index data is stored for each of the devices, and wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space; A document index search unit that, at the same time that the query confirmation unit confirms that the query has been recorded, reads the contents of the query and searches a document index database for a document having a word group that matches the query, When the document index search unit searches and extracts document information matching the user query, the shared memory A registration query processing unit having a search result recording unit for recording the document information therebetween, and a search result for constantly checking whether or not the document information recorded by the search result recording unit is recorded in the shared memory space. A confirmation unit and the search result confirmation unit, when appropriate document information is described as a search result for the user query,
A search result processing unit having a search result output unit that reads out the document information that is a search result recorded in the shared memory space and outputs the read document information to a user.

According to a fifth aspect of the present invention, there is provided a document search system in which a plurality of distributed document search devices share a shared memory space to perform a document search, wherein the document search system includes a plurality of document search devices. A document index database in which different document index data is stored for each of the document search devices, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space; A document index search unit that reads the contents of the query and searches the document index database for a document that has a word group that matches the query at the same time that the query unit confirms that the query has been recorded. When the document index search unit searches and extracts document information that matches the user query, A registration query processing unit having a search result recording unit that records the document information in the memory space; and a search that constantly checks whether or not the document information recorded by the search result recording unit is recorded in the shared memory space. A result confirming unit and the search result confirming unit, when appropriate document information is described as a search result for the user query, reads out the document information that is the search result recorded in the shared memory space, And a search result processing unit having a search result output unit that outputs the search results. Therefore, each of the document search apparatuses constituting the document search system can perform the search processing independently and can merge the search results at high speed.
This makes the document index stored in the document index database of each search device different,
That is, the document index is created in a stripe shape.

According to a sixth aspect of the present invention, there is provided a document search system in which a plurality of distributed document search devices share a shared memory space to perform a document search. A document index database in which the same document index data is stored for each of the devices, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space; And a document index search unit that, when the query confirmation unit confirms that the query has been recorded, reads the contents of the query and searches for a document having a word group that matches the query from the document index database. When each of the other distributed document search apparatuses records a search result in the shared memory space, A search result recording unit that confirms that the document index search unit does not record the search result, and prevents the document index search unit in each of the other document search devices from recording the search result during recording. A registration query processing unit, a search result confirmation unit that constantly confirms whether or not the document information recorded by the search result recording unit is recorded in the shared memory space, and the search result confirmation unit responds to the user query. When appropriate document information is described as a search result, the relevant document information that is a search result recorded in the shared memory space is read out,
And a search result processing unit having a search result output unit for outputting to a user.

According to a sixth aspect of the present invention, there is provided a document search system in which a plurality of distributed document search devices perform a document search by sharing a shared memory space. Each of the plurality of document search devices has a document index database in which the same document index data is stored, and each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space. A query confirmation unit, and a document index retrieval unit that reads the contents of the query at the same time that the query confirmation unit confirms that the query has been recorded, and retrieves a document having a word group that matches the query from the document index database. Unit, when each of the other document search apparatuses distributed records a search result in the shared memory space, each of the other document search apparatuses A search result recording unit that confirms that the document index search unit does not record the search result, and prevents the document index search unit in each of the other document search devices from recording the search result during recording. A registration query processing unit having a search result confirmation unit that constantly checks whether or not the document information recorded by the search result recording unit is recorded in the shared memory space; and the search result confirmation unit includes the user query. A search result processing unit having a search result output unit that reads out the relevant document information that is a search result recorded in the shared memory space when appropriate document information is described as a search result for the Of the document search devices that make up the document search system, use the search result that responds the fastest and respond the fastest. It can be fused to the search results. This is because the document indexes stored in the document index database of each search device are the same, that is, the document indexes are mirrored.

According to a seventh aspect of the present invention, a user query is input, and the input user query is morphologically analyzed using a morphological analysis dictionary to determine a word described in the user query. In addition, the part of speech of each word, the inflected form, and the like are determined, and a user query process for recording the morphologically analyzed query in a shared memory space, a document to be processed is input, and a morphological analysis dictionary is used for the input document. Morphological analysis to determine the words of each sentence described in the document, and to record the words and document names contained in each sentence as an index of each document in a document index database; Constantly checking whether there is a query stored in the space, and confirming that the query has been recorded in the shared memory space; Read the contents of the query from space,
A document having a word group that matches the query is searched from the document index database, and when extracting document information matching the user query, a registration query process of recording the document information in the shared memory space; It is constantly checked whether or not the document information recorded by the registration query processing is recorded in the shared memory space, and when appropriate document information is described as a search result for the user query, the document information is recorded in the shared memory space. And a computer-readable recording program for causing a computer to execute a document search method characterized by comprising: a search result process that reads out the document information that is the searched result and outputs the document information to a user. The medium is the gist.

According to the seventh aspect of the present invention, a user query is input, and the input user query is morphologically analyzed using a morphological analysis dictionary to determine a word described in the user query. A user query process for determining the part of speech, inflected form, etc. of each word, and recording the morphologically analyzed query in a shared memory space; inputting a document to be processed; and a morphological analysis dictionary for the input document. Morphological analysis using, to determine the word of each sentence described in the document, and the target document processing to record the word and the document name included in each sentence as a document index in a document index database, It is constantly checked whether there is a query stored in the shared memory space, and at the same time it is confirmed that a query has been recorded in the shared memory space, When the content of the query is read from the memory space, a document having a word group that matches the query is searched from the document index database, and when the document information suitable for the user query is extracted, the document information is stored in the shared memory space. In the registration query process of recording the document information, it is constantly checked whether or not the document information recorded by the registration query process is recorded in the shared memory space, and appropriate document information is described as a search result for the user query. Read out the document information, which is the search result recorded in the shared memory space, and output the search result to the user. This is recorded on a computer-readable recording medium. To read a program recorded in the recording medium into a computer system,
By executing, a large amount of document indexes and a large number of user queries can be scalably processed.

[0018] The invention according to claim 8 has a document index database in which different document indexes are stored for each of a plurality of distributed document search devices, and the plurality of document search devices share a shared memory space. A document search method in a document search system that performs a document search by using a plurality of document search devices, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space. At the same time as confirming that the query has been recorded, the content of the query is read from the shared memory space, a document having a word group that matches the query is searched from the document index database, and document information matching the user query is obtained. When extracting the shared information, a registration query process for recording the document information in the shared memory space; It is constantly checked whether or not the document information recorded by the registration query process is recorded in the memory space, and when appropriate document information is described as a search result for the user query, the document information is recorded in the shared memory space. And a search result process for reading the document information as the search result and outputting the document information to the user. The computer reads a computer program for executing the document search method in the document search system. The possible recording medium is the gist.

According to the eighth aspect of the present invention, there is provided a document index database in which different document indexes are stored for each of a plurality of distributed document search devices, and the plurality of document search devices occupy a shared memory space. A document search method in a document search system that performs a shared document search, wherein in each of the plurality of document search devices,
It is constantly checked whether or not there is a query stored in the shared memory space, and at the same time it is confirmed that a query has been recorded in the shared memory space, the content of the query is read from the shared memory space, and the query and A document having a matching word group is searched from a document index database, and when extracting document information matching the user query, a registration query process of recording the document information in the shared memory space; and It is constantly checked whether or not the document information recorded by the registration query process is recorded, and when appropriate document information is described as a search result for the user query, the search result recorded in the shared memory space And a search result process for reading out the document information and outputting it to the user. Since the program for causing the computer to execute the document search method in the search system is recorded on a computer-readable recording medium, the program recorded on the recording medium is read into a computer system, and executed. Each document search device constituting the document search system performs a search process independently, and the search results can be merged at high speed. This is because the document indexes stored in the document index database of each search device are different, that is, the document indexes are created in a stripe shape.

According to a ninth aspect of the present invention, there is provided a document index database in which the same document index is stored for each of a plurality of distributed document search devices, and the plurality of document search devices share a shared memory space. A document search method in a document search system for performing a document search by continuously checking whether or not there is a query stored in the shared memory space in each of the plurality of document search devices. At the same time as confirming that the query has been recorded, the contents of the query are read from the shared memory space, a document having a word group that matches the query is searched from the document index database, and each other distributed document is searched. When the search device records the search result in the shared memory space, each of the other document search devices does not record the search result. And whether the document information recorded by the registration query process is recorded in the shared memory space so that each of the other document retrieval devices does not record a search result during recording. Whether the document information is described as a search result for the user query and appropriate document information is described as a search result for the user query, the search result is output to the user by reading the document information, which is the search result recorded in the shared memory space. And a computer-readable recording medium that records a program for causing a computer to execute the document search method in the document search system, which has a result process.

According to the ninth aspect of the present invention, there is provided a document index database in which the same document index is stored for each of a plurality of distributed document search devices, and the plurality of document search devices are provided in a shared memory space. A document search method in a document search system for performing a document search by sharing a document, wherein each of the plurality of document search devices includes:
It is constantly checked whether or not there is a query stored in the shared memory space, and at the same time it is confirmed that a query has been recorded in the shared memory space, the content of the query is read from the shared memory space, and the query and A document having a matching word group is searched from the document index database, and when each of the other distributed document search devices records the search result in the shared memory space, each of the other document search devices records the search result. The registration query processing for confirming that the search has not been performed, and preventing the other document search apparatuses from recording search results during recording, and the document information recorded by the registration query processing in the shared memory space. It is constantly checked whether or not it is recorded, and when appropriate document information is described as a search result for the user query, A computer-readable program for causing a computer to execute a document search method in a document search system, comprising: a search result process that reads out the document information that is a recorded search result and outputs the document information to a user. Since the program is recorded on a simple recording medium, the program recorded on the recording medium is read into a computer system and executed, whereby the fastest response can be achieved among the document retrieval apparatuses constituting the document retrieval system. Utilizing search results, it is possible to fuse search results that respond fastest. This is because the document indexes stored in the document index database of each search device are the same, that is, the document indexes are mirrored.

[0022]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows the overall configuration of a document search device according to an embodiment of the present invention. In the figure,
The document search device stores a shared memory S-1, a morphological analysis dictionary D-1, a processing target document database D-2 in which a processing target document is stored, and a document index for specifying the document. A document index database D-3.

The document search device further includes a user query processing unit A-1 for inputting a user query, morphologically analyzing the input user query, and recording the analysis result in a memory space of the shared memory S-1. Enter the document to be processed,
A morphological analysis is performed on the input document to create a document index, and a target document processing unit A-2 that records the document index in the document index database D-3 and the presence or absence of a recorded query in the shared memory space are checked. When a query is recorded in the memory space of No. 1 and a document having a word group that matches the query is searched from the document index database D-3, and when the document information suitable for the user query is extracted, The presence / absence of document information recorded by the registration query processing unit A-3 for recording the document information in the memory space of the memory S-1 and the registration query processing unit A-3 in the memory space of the shared memory S-1 is confirmed. When appropriate document information is described as a search result for the user query, the document information is recorded in the memory space of the shared memory S-1. Reads the document information is a search result, the search result processing section A-4 to be output to the user
And

The user query processing unit A-1 includes a user query input unit T-1 for inputting a user's search query (referred to as a user query), and a user query input by the user query input unit T-1 for morphological analysis. A morphological analysis is performed using the dictionary D-1, the input query is divided into words, the words described in the user query are determined, and grammatical information such as the part of speech and inflected forms of each word is determined. Query morphological analyzer T-2 to be determined, Query morphological analyzer T-2
Has a grammatically analyzed user query grammatical information and a query recording unit T-3 for recording the user query in the memory space of the shared memory S-1.

The target document processing unit A-2 includes a target document input unit T-40 for inputting a document to be processed, and a morphological analysis dictionary D-
1 to determine the words of each sentence described in the document, and to record the words and document names included in each sentence as a document index in the document index database D-3. And a creation unit T-5.

Further, the registered query processing section A-3 includes a query confirmation section T-6 for constantly confirming whether or not there is a query recorded in the memory space of the shared memory S-, and a query confirmation section T.
-6 confirmed that the query was recorded,
The content of the query is read, and a document having a word group that matches the query is stored in a document index database D-3
When the document index search unit T-7 and the document index search unit T-7 search from and extract the document information matching the user query, the document information is recorded in the memory space of the shared memory S-1. Search result recording section T
-8.

The search result processing section A-4 includes a shared memory S-
1. A search result confirming unit T-9 for continuously confirming whether or not the document information recorded by the search result recording unit T-8 is recorded in the memory space of No. 1; When appropriate document information is described as a search result for the search result, the search result output unit T-10 reads out the relevant document information, which is the search result recorded in the memory space of the shared memory S-1, and outputs it to the user. And

Next, as an example of a document search device according to an embodiment of the present invention, a document search device applied to the Internet will be described. First, the user query input unit T-1
Inputs a user's search query. For example, if the user enters a query "Today's weather" by the input means,
The user query input unit T-1 inputs the input character string and converts the input character string into a user query morphological analysis unit T that performs the following processing.
-2.

The user query morphological analysis unit T-2 morphologically analyzes the search query input by the user into the user query input unit T-1 using the morphological analysis dictionary D-1, and converts the input query into words. Divide into words and analyze grammatical information such as part of speech and inflected forms for each word. Here, the morphological analysis refers to searching a word dictionary for an input character string, and performing part-of-speech information (part-of-speech), head-of-sentence information (head-of-sentence possible), forward connection information (preceding), backward connection information ( Acquisition of grammatical information, such as postscript).

In an ordinary word dictionary, a high-speed search can be performed by performing a special dictionary structure called a TREI dictionary structure. If the dictionary entry includes "Ah", "Aiatsu", "Ai", etc., each of the first characters (in this case, Japanese characters, 2 bytes of Japanese characters
), The second letter is the same, etc.
Each of them is sequentially configured in a tree structure. When the last character is matched, information such as part of speech information (part of speech), sentence availability information (heading possible), forward connection information (preceding), backward connecting information (postfixed), etc. for the word dictionary item Is described.

The head-of-sentence availability information is a flag indicating whether or not it can be at the head of the sentence. If the beginning of the sentence is acceptable, the word may be present at the beginning of the sentence, but if the sentence is not the beginning of the sentence, the word is not permitted to be at the beginning of the sentence. The forward connection information is permitted to be connected only when the part of speech or attribute of the preceding word is appropriate, and is deleted as a candidate when the preceding word is not permitted to be connected. Similarly, the backward connection information is permitted to be connected only when the part of speech or attribute of the subsequent word is appropriate, and is deleted as a candidate when the word is not permitted to be connected later.

A candidate is selected by such a part-of-speech connection. The maximum likelihood candidate is selected by a method called a minimum cost method. The minimum cost method is a processing method in which a morpheme candidate having the lowest cost is set as a maximum likelihood candidate. There are two types of costs used in morphological analysis: (1) connection cost and (2) word cost. Of these, the connection cost is the cost required to connect a certain word to another word. Since it is a word and a word, the connection cost for the word + the utilization is zero. The word cost is a cost related to the word. For example, a frequently used word has a low cost. Also, since utilization is not a word, the cost is zero. By the morphological analysis, the text part is decomposed into words, and at the same time, parts of speech that are considered to be correct are given to each word.

In the above example, the morphological analysis is performed as follows.
You.  Notation part of speech independent word  How is the weather today?  cost = 49  Today's noun ○ particle is a weather noun ○ is a particle? Question mark

The user query is divided into words by morphological analysis, and grammatical characteristics such as part of speech and inflected forms of each word are extracted. In general, a document search device on the Internet or the like often extracts only independent words from information that has been subjected to morphological analysis and uses the extracted words as queries. In the case of the query "What is the weather today?", The independent words are "Today" and "Weather" from the morphologically analyzed grammatical information. The query recording unit T-3 which performs the following processing on this information
Pass to.

The query recording unit T-3 records the query in the memory space of the shared memory S-1 together with the grammatical information of the query analyzed by the user query morphological analysis unit T-2. For queries, describe each query in a specific area. As the query, the query input by the user and the word-by-word information (“Today” and “Weather” in the above example) obtained by morphologically analyzing the query are recorded in the memory space of the shared memory S-1.

Next, the target document input unit T-4 outputs the search target
From the document database D-2 to be processed
I do. For example, suppose we have the following set of documents:
It is.  Document number Document contents  Document 1: Today's weather in Tokyo is rain. The weather is fine tomorrow. Document 2: The average temperature in Tokyo is 20 degrees Celsius, followed by hot and humid days. Document 3: The population of the Internet in Tokyo is about 500,000. …  Enter all of these documents.

The document index creation unit T-5 morphologically analyzes the input document using the morphological analysis dictionary D-1 to search for the document input by the target document input unit T-4. The sentence in the document is divided into words, and grammatical information such as part of speech and inflected forms is analyzed for each word,
The characteristics of each document are recorded as a document index in the document index database D-3.

The method of performing the morphological analysis is the same as that of the user query morphological analysis unit T-2. In particular, in order to output the search result appropriately, the same morphological analysis process and the same morphological analysis dictionary D-1 are not used, and the result of the morphological analysis differs between the query and the document index. In the user query morphological analysis unit T-2,
During the morphological analysis, independent words in the query are extracted. Therefore, the document index creation unit T-5 similarly performs a morphological analysis on the document to be processed in the morphological analysis dictionary D-1 and further extracts an independent word.

In the above example, it is determined that the word is an independent word in the document contents.
Words are underlined.  Document number Document Naito  Document 1:TokyooftodayofweatherIsrain.tomorrowofweatherIsSunny. Document 2:TokyoofAverage temperatureIs20 degreesWhenSultry dayButit continues. Document 3:Tokyoofthe InternetofpopulationIsAbout 500,000It is.… In the previous example, the independent words extracted are as follows.  Document number Independent word  Document 1: Tokyo, today, weather, rain, tomorrow, weather, sunny Document 2: Tokyo, average temperature, 20 degrees, sultry day, followed Document 3: Tokyo, Internet, population, about 500,000… In the document index creation unit T-5, these document numbers
And a pair of independent words as a document index
In the database D-3.

The query confirmation unit T-6 is provided in the shared memory S-1.
It is confirmed whether or not there is a new query recorded by the query recording unit T-3 in the memory space of. Since a plurality of queries are recorded in the memory space of the shared memory S-1, the query checking unit T-6 checks the next query if there is no new query. If there are no new queries, search for the next query. Only when there is a new query, the processing is passed to the document index search unit T-7. For example, if a query such as "What is the weather today?" (Independent words, "Today", "Weather") is written in the area of QID (query ID) = 1, the query confirmation unit T-6 Is QID = 1
Is checked, it is confirmed that a new query has occurred, and the process proceeds to the document index search unit T-7. Where QI
D means the ID of the query described in the memory space of the shared memory S-1, and is equal to the address on the shared memory.

When a new query is found in the memory space of the shared memory S-1 by the query check unit T-6, the document index search unit T-7 searches the document index stored in the document index database D-3. A search is performed, and a document set most suitable for the query is output as a search result. For example, assuming that the query "What is the weather today?" When the area of the document is inspected, it is confirmed that a new query has occurred, and the document index database D
-3 and returns the document number that has an independent word that matches the independent word of the query. In this example, since “today” and “weather” completely match in document 1, document 1 is returned as a search result.

The search result recording unit T-8 records a document set suitable for the new query created by the document index search unit T-7 as a search result in the memory space of the shared memory S-1. In the above example, it is understood that the number of conforming documents is 1 and the conforming document name is “document 1” as the search result.
In addition, to indicate the degree of conformance of each conforming document,
The relevance is usually determined, and the search results are recorded in the memory space of the shared memory S-1 in the order of the most suitable according to the relevance. A method based on the tf * idf method can be considered as a scale indicating the degree of matching in the search processing. This is an equation that can calculate a high degree of relevance when the word that is most important matches from the number of words of the matched query in the entire document and the weight of words in the matched document.

The search result confirming unit T-9 includes a shared memory S-
It is confirmed whether or not there is a new search result recorded by the search result recording unit T-8 in the first memory space. Shared memory S-
In one memory space, there are a plurality of search results as well as a plurality of queries. It is checked whether there is a newly searched result among these search results. If there is no search result, the area of the next search result in the memory space of the shared memory S-1 is confirmed.

The search result output unit T-10 reads the new search result confirmed by the search result confirmation unit T-9 from the memory space of the shared memory S-1 and outputs it as a search result. In the above example, the number of conforming documents is 1 and the conforming document name is “document 1” as the search result.

FIG. 2 shows the processing contents of the query recording unit T-3 in the user query processing unit A-1. In the figure, a related library is loaded in step S11,
Next, the shared memory S-1 is made accessible (S
-12). Further, initialization is performed in step S13, and then the QID for checking the free area of the shared memory S-1 describing the query is set to QID = 0 (step S14).

In step S15, the query area of the shared memory S-1 specified by the QID is checked, and in step S16, it is determined whether the query area of the shared memory S-1 specified by the QID is free. . Here, the query area refers to an area on the shared memory S-1 that describes a query received by each search server. Step S1
If the determination in step 6 is affirmative, Q is determined in step S17.
The query area of the shared memory S-1 specified by the ID is initialized, a query is written, and the process ends.

On the other hand, if the determination in step S16 is negative, the QID is incremented in step 18, and
In the next step S19, it is determined whether or not the value of the QID is larger than the maximum number of query areas. Here, the maximum number of query areas means the total number of areas in the query area where the query is written. If the determination in step 19 is affirmative, that is, if it is determined that the entire query area is not empty, the process waits for a predetermined time until the query area becomes empty (step S20), and the process returns to step 14. , QID = 0, the available query area is checked.

If the determination in step S19 is denied, that is, if it is determined that all available query areas have not been checked, the process returns to step S15, and the QI of the value incremented in step S18.
It is checked whether there is a free space in the query area of D. Thereafter, the same processing is repeated.

Next, FIGS. 3 and 4 show the processing contents of the query confirmation section T-6, document index search section T-7 and search result recording section T-8 in the document search apparatus shown in FIG. 3 and FIG. 4, steps S21 to S27 are processes of the query confirmation unit T-6, and steps S28 and S29 are performed.
Is a process of the document index search unit T-7, and step 30 is a process of the search result record T-8. In these figures, first, the query confirmation unit T-6 makes an initial setting in step S21, and makes the shared memory S-1 accessible (step S22).

Next, in step S23, the QID is set to QID =
The query area is designated as 0, and the query area of the shared memory S-1 designated as QID = 0 is checked in step S24. Next, at step 25, it is determined whether or not a query has been written in the query area of QID = 0. If the determination in step S25 is negative, the value of the QID is incremented in step S26, and then it is determined in step S27 whether the value of the QID is greater than the maximum number of query areas.

If the determination in step 27 is affirmative, that is, if it is determined that the check of whether or not the query has been recorded has been completed for all of the query areas, the process returns to step 23, where QID = 0 and Initiate a check to see if a query has been recorded.

If the determination in step S27 is denied, that is, if it is determined that the check for the presence or absence of the query has not been completed for all of the query areas, the process returns to step S24, and the value is incremented in step S26. Check the query area for the value QID. Further, if the determination in step S25 is affirmative, that is, if it is determined that the query has been written in the query area of the specified QID, the query confirmation unit T-6 sends the processing to the document index search unit T-7. hand over.

The document index search unit T-7 obtains a query in step 28, searches the document index database D-3 according to the query obtained in step S29, and sets a document set most suitable for the query as a search result. Output. Here, the document index search unit T-7 passes the processing to the search result recording unit T-8. The search result recording unit T-8 records the search result output by the document index search unit T-7 in the memory space of the shared memory S-1, and notifies the query area designated by the QID that the search has been completed. The search completion flag is described, and the process is passed to the query confirmation unit T-6. Accordingly, the query confirmation unit T-6 repeats the same processing as described above from step S24.

As described above, the query confirmation unit T-6 constantly confirms whether or not a query is recorded in all the query areas of the shared memory S-1, and a new query is stored in the memory space of the shared memory S-1. Is found, the document index search unit T-7 searches the document index database D-3 and outputs a search result, and the search result recording unit T-8 stores the search result in the memory of the shared memory S-1. Record in space.

Next, FIG. 5 shows the processing contents of the search result confirmation section T-9 and the search result output section T-10 of the document search apparatus shown in FIG. In the figure, steps S31 to S35 are processes of the search result confirmation unit T-9, and steps S36 and S37 are processes of the search result output unit T-10. In FIG.
The search result confirmation unit T-9 initializes the query register QID in step S31, and then reads data in the query area specified by the QID in the shared memory S-1 in step 32. In step 33, it is determined whether or not search end flags from all computers functioning as search servers have been written in the query area specified by the QID in the shared memory S-1.

If the determination in step 33 is negative, it is determined in step 34 whether a predetermined time has elapsed. If it is determined in step 34 that the predetermined time has not elapsed, the process returns to step 32, and the same processing as described above is performed. If the determination in step 33 is affirmative, that is, if it is determined that the search end flags from all the computers functioning as search servers have been written in the query area specified by the QID in the shared memory S-1 If it is determined in step S34 that the predetermined time has elapsed, the process proceeds to step S35, and the process proceeds to step S35.
Reads the search result from the shared memory S-1. Here, the processing is performed from the search result confirmation unit T-9 to the search result output unit T-10.
Move to

In the search result output unit T-10, step S
In step S36, the search result confirmation unit T-9 proceeds to step S35
And outputs the search result read in step S37, initializes the query area in the shared memory S-1 in step S37, terminates the access to the shared memory S-1, and ends this processing.

Next, FIG. 6 shows a configuration of an embodiment in which the document search device according to the embodiment of the present invention is applied to a document search system on the Internet. 6, the document search system includes a plurality of HTTP servers C-1, a document processing server C-2, a plurality of search servers C-3, a morphological analysis dictionary D-1, and a processing target database D. -2, a document index database D-3, and a shared memory S-1.

Each HTTP server C-1 has the user query processing section A-1 and the search result processing section A-4 shown in FIG. It has functions of an analysis unit T-2, a query recording unit T-3, a search result confirmation unit T-9, and a search result output unit T-10. In a document retrieval system on the Internet,
A web server is an engine part that processes user queries and user requests. As functions of the WEB server, the functions of T-1, T-2, T-3, T-9 and T-10 are provided by a CGI (common gateway interface) on the WEB server side.
ce) Equivalent processing. The server that performs the CGI service is the HTTP server C-1.

The document processing server C-2 has the target document processing section A-2 shown in FIG. 1, and has the functions of the target document input section T-4 and the document index creation section T-5 described above. ing. Further, each search server (TRET server) C-3 has the registered query processing unit A-3 shown in FIG. 1, and the above-described query confirmation unit T-6, document index search unit T-7, and search result recording unit It has the function of T-8. The morphological analysis dictionary D-1 and the processing target document database D-
2. The document index database D-3 is basically the same as that shown in FIG. However, FIG.
Although not shown, the document index database D-3 shown in FIG. 3 is provided for each of the plurality of search servers C-3.

In this embodiment, the HTTP server C-1 and the search server C-3 can be accessed in parallel using the special setting register and the shared memory S-1. Search server C-
3 basically returns a document segment having a matching keyword in its own index as a result. In that case, it is possible to return from the highest score. Various services performed by the HTTP server C-1, the document processing server C-2, and the search server C-3 are performed via the shared memory S-1. Here, the shared memory means that each of the servers C-1, C-2, and C-3 is connected by another network, and different processes (CPUs) of the respective servers C-1, C-2, and C-3 are performed via the CPUs. process) or thread (T
hread) is memory that can be shared.

FIG. 7 shows the configuration of a storage area provided for processing a user query in the shared memory S-1. In the figure, a storage area provided on the shared memory S-1 for processing a user query input from the user to the user query processing unit A-1 of the HTTP server C-1 is stored in each search server (TRET server). ) Has a flag area in which a flag group indicating that the processing for the user query has been completed is recorded, a query area, and a search result area. The flag area is provided with a bit for recording a processing end flag indicating that the search processing has been completed for each search server. In the initial state, the bit indicating each processing end flag is set to “0”. When the processing of each search server ends, the bit in which the processing end flag corresponding to the search server is recorded is set to “1”. HTT
The P server C-1 determines that the search processing of all search servers has been completed when all the processing end flags have become “1”. In this embodiment, 16 bytes are allocated to this flag area.

The query area is an area in the memory space of the shared memory S-1 that describes a query received by each search server. Generally, a plurality of queries can be described in the query area. In this embodiment, the query area is allocated as a data area for describing the contents of the query from the HTTP server C-1. The search result area is a data area in which the first to n-th search results for each query from each search server C-3 are written.
For example, the data area used to write each search result
(32 × n) bytes are allocated as 32 bytes. When writing a search result in the search result area, the data is sorted and written according to a score that is the degree of matching between the query and the search result.

That is, each search server C-3 describes the search results in the search result area while sorting, when its own thread has the search results. However,
Lock the record when searching while sorting so that it does not go wrong during sorting. (The search server determines whether or not writing is possible according to the processing end flag of each search server.)

The HTTP server C-1 has a shared memory S-
1, a user query is described, and each search server C-3 returns an answer to the query, or after a certain period of time, returns the query result together to the Internet. On the other hand, if there is a query in the memory space of the shared memory S-1, the search server C-3 immediately processes the query. When outputting the search result, each search server C-3 sorts and collects only those having the highest score with respect to the data in the memory space of the shared memory S-1.
After outputting its own search result, the search server whose processing is the last describes to the HTTP server C-1 that all the search results have been output in the special setting register. (HT
The TP server C-1 checks the value of this special setting register at certain time intervals.)

If there is a possibility that the same address of the shared memory S-1 may be updated by each node, it is necessary to perform an interrupt or perform some processing. That is, here, the search results are written out while each search server performs C-3 sorting. Therefore, the area for describing the search result is small, and only a fixed number of search results need to be output. On the other hand, in order to repeat the sorting, the update must not be performed when another search server is performing the sorting.

The embodiment shows a case where the shared memory S-1 is 128 KB. FIGS. 8 to 11 are flowcharts showing specific processing contents in the above-described document search system. These correspond to FIGS. 2 to 5, and the flowcharts of FIGS. Since this is a replacement, the details of the description of each flowchart are omitted, and only a general description is given. 8 shows the processing of the query recording unit T-3 in FIG. 2, and FIGS. 9 and 10 show the query confirmation unit T-6, the document index search unit T-7, and the search result recording unit T-8 in FIGS. 11 is shown in FIG.
Of the search result confirmation unit T-9 and the search result output unit T-10.

The search result recording section T-8 in FIG.
FIG. 12 shows an example of the details of the processing of step 30 'by (search server C-3), and FIGS. 13 and 14 show other examples of the details of the processing of step 30'. Here, FIG. 14 is a flowchart showing the processing content of the flow A in FIG. 8 to 11, when the document index database D-3 is composed of a plurality of documents, there are a plurality of composition methods. Level 0 is a stripe, level 1 is mirroring, and level 2 is a composite configuration of stripe and mirroring. This stripe does not have a document index database D-3 having the same document index.
-3 document index database D provided for each
3 is distributed.

On the other hand, "mirroring" means that there is a document index database D-3 which has the same document index. On the other hand, the composite configuration of stripe and mirroring indicates that there is a mirrored document index and a non-mirrored document index. When a document index is stored in the document index database D-3 provided for each of the plurality of search servers C-3, the stripe means that the plurality of search servers C-3 do not have the same document index, but are different from each other. Is stored in the document index D-3 provided for each search server C-3.

After describing the query on the shared memory S-1, each search server C-3 sets the query confirmation unit T-
At 6, the query is fetched and based on the document index,
Using a document index database D-3 managed by itself, a search is performed by a document index search unit T-7,
The search result is recorded in the search result recording unit T-8. In the striped indexing method, the search results are stored by combining the search results from a plurality of search servers C-3.

The output contents of the search result include “conforming document name (filePath)”, “conforming document score (score)”,
“Relevance order”, “ID when creating document index (segmen
t-ID) "are described. These search results are the results in the respective document indexes, and if these are not combined with the results from all the plurality of search servers C-3, they are simply striped. Only the result of a partial document fragment is output, not the result for the entire document set.

As inputs to the document index search unit T-7, results of a "search query" and a "matching document score area (n-th to + m-rank scores of documents matching the search query)" are output. For example, if you set n = 1 and m = 40, you can output the results from the 1st to the 41st place in the score. In addition, it is also possible to output “the score from the nth rank to the + mth rank among the documents that match the search query.” This is also described in the shared memory S-1 similarly to the ordinary query. Will be described in detail based on the processing flow of the striped search result.

When two sets of stripes (that is, A, B
Search servers (C-3) composed of two document sets
A, C-3B). Since there are two sets of stripes, consider a simple configuration of two sets of documents to be processed and two search servers for further processing. Since there are two sets of documents to be processed, the document index creation unit T-5 for creating an index of the documents to be processed also creates two sets of document indexes.
(Respectively IA and IB). Document indexes IA and IB
Are stored in the document index search unit T-7 on another search server. The user inputs a search query, and the query confirmation unit T-6 on each search server
The query is confirmed, and the document index is searched using the document indexes (IA, IB) on the respective search servers. It is assumed that the search server C-3A outputs the search result RIA using the document index IA.

To record the search result RIA in the designated area of the shared memory S-1, the search result recording unit T-8 first stores the designated area in the shared memory S-1 with another search server (here, Then, it is confirmed that C-3B) has not been written. (This is because the shared memory S-
The designated area in 1 is for the search server that records all search results to use the same area. That is,
For a certain search query, the area where search results are recorded is 1
It will be prepared in some places. If another search server is writing, wait until writing is completed. On the other hand, when the other computer has not performed the writing, the search result RIA is written on the shared memory S-1.

When writing to the shared memory S-1,
When writing search results at the very beginning, search result RIA
COPY directly. On the other hand, if you want to overwrite the result already written by another search server instead of the first one, sort the result written by another search server according to the SCORE of the search result While doing
The one with the largest SCORE is stored at the top. FIG. 12 shows the processing flow.

Next, to obtain the next result of the maximum number of display results (for example, 40), the following processing is performed. On the shared memory S-1, the query record and the matching document score area are stored in the query recording unit T-3 after the maximum display result number (that is, the display order from the maximum display result number to the maximum display number * 2). Set to return results. In a normal method, when the results output by a plurality of units are sorted with respect to the maximum number of display results and output as the final result, it is necessary to sort them all. In this case, as the number of display results increases, the time required for the sorting process is required, and the processing speed is reduced.

On the other hand, in the document search device and the document search system according to the embodiment of the present invention, the history of the search results is stored in the shared memory S-1. On the shared memory S-1, all of the shared memories S-
Until the content of No. 1 is overwritten, the history of search results is accumulated in the shared memory, but no more than that, and no history of search results remains. By using such a history of search results, it is possible to sort at high speed and output the search results. In other words, it uses search results before a certain display order, does not perform processing before the display order,
Only the search results after the rank to be displayed need to be processed. In this case, for example, if the display order is from the 1000th place, the sort processing is not required at all for the results before the 1000th place, and the processing speed is increased. 13 and 14 show the processing contents. Note that the order of the search results to be displayed is from the X-th to the (X + Y) -th.

Further, when search results after the maximum display result number * 2 are required, the same processing is performed for the maximum display result number * 2.
Do for In the case of the stripe format, the search result output unit T-10 which outputs the search result last (this is usually
As described with reference to FIG. 5 and FIG. 11, if the HTTP server C-1 does not output the search result for the search query within a predetermined time,
Until now, each search server C recorded in the shared memory S-1
The search result recorded by the search result recording unit T-8 is output through the search result output unit T-10.

In the case of a mirror, the same processing is performed for the same search query by a plurality of search servers for the same document index. That is, for the same search query, the same search result is output by a plurality of search servers. Of course, since it is a mirror, it usually has the same search result. However, in order to enhance reliability, it is also possible to confirm whether a plurality of search servers output the same search result.

On the other hand, the meaning of a normal mirror is to improve the search speed. That is, in a search server having the same mirror, the time from when a search query is given to when a search result is obtained is not the same. Therefore, the feature that retrieves the result of outputting the search result earliest to the computer that has the document index of the same mirror,
This is a search result confirmation unit T-9. In the processing flow of the query recording unit T-3, the search result confirmation unit T-9, and the search result output unit T-10 shown in FIGS. 8 and 11, If (query-reg
If the document index is IA or IB and the mirror is MIA or MIB in the part where ister [QID] == 0x0f), either IA or MIA stores the search result on the shared memory. If the search result is described in the shared memory by either MIB or IB, query-register
By setting [QID] == 0x0f, processing at the mirror is realized. That is, the earlier search result of the document index IA or MIA can be output as the search result in the search result output unit T-10. Similarly, it is possible to output a search result that is earlier than the document index IB or MIB.

A user's query is input, and the input user query is morphologically analyzed using a morphological analysis dictionary to determine the words described in the user query. User query processing for determining the shape and the like and recording the morphologically analyzed query in the shared memory space, inputting a document to be processed, morphologically analyzing the input document using a morphological analysis dictionary, and The target document processing of determining the words of each sentence described in the above, and recording the words and the document names included in each sentence as an index of each document in the document index database, and the query stored in the shared memory space. It is constantly checked whether or not there is a query. Registration query processing for retrieving a document having a word group that matches the query from the document index database and extracting document information matching the user query when the document information is extracted in the shared memory space And constantly check whether or not the document information recorded by the registration query processing is recorded in the shared memory space, and when appropriate document information is described as a search result for the user query, the shared memory A computer-readable recording program for causing a computer to execute a document search method characterized by having a search result process for reading the document information as search results recorded in a space and outputting the document information to a user It may be recorded on a medium. By causing the computer system to read and execute the program recorded on the recording medium, a large amount of document indexes and a large number of user queries can be scalably processed.

A document index database in which different document indexes are stored for each of a plurality of distributed document search devices, wherein the plurality of document search devices share a shared memory space to perform a document search. A document search method in a system, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space, and determines that the query is recorded in the shared memory space. At the same time as confirming, read the contents of the query from the shared memory space,
When a document having a word group that matches the query is searched from a document index database and document information that matches the user query is extracted, a registration query process for recording the document information in the shared memory space; It is constantly checked whether or not the document information recorded by the registration query process is recorded in the memory space, and when appropriate document information is described as a search result for the user query, the document information is recorded in the shared memory space. A computer-readable recording program for causing a computer to execute a document search method in a document search system, comprising: a search result process for reading the document information as a search result and outputting the document information to a user. It may be recorded on a medium. By causing the computer system to read and execute the program recorded on the recording medium, each document search device constituting the document search system performs a search process independently, and the search results can be merged at high speed.

A document index database in which the same document index is stored for each of a plurality of distributed document search devices, wherein the plurality of document search devices share a shared memory space to perform document search. A document search method in a search system, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space, and the query is recorded in the shared memory space. At the same time, read the contents of the query from the shared memory space,
A document having a word group that matches the query is searched for from the document index database, and when each of the other distributed document search devices records a search result in the shared memory space, each of the other document search devices performs a search. Confirming that the result has not been recorded, and a registration query process in which each of the other document search devices does not record a search result in the middle of recording, and a registration query process recorded in the shared memory space by the registration query process. Always check whether or not the document information is recorded, as a search result for the user query, if appropriate document information is described, read the document information is a search result recorded in the shared memory space, Causing a computer to execute a document search method in a document search system having a search result process for outputting to a user. Programs may be recorded in a recording medium readable by a computer for. By causing the computer system to read and execute the program recorded on the recording medium, among the document retrieval devices constituting the document retrieval system, the retrieval result that responds at the highest speed is used, and the computer responds at the highest speed. Search results can be merged.

Note that the “computer system” here includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” is a floppy disk, a magneto-optical disk, an RO
M means a storage medium such as a general medium such as a CD-ROM or a hard disk built in a computer system. Further, a "computer-readable recording medium" refers to a communication line for transmitting a program via a network such as the Internet or a communication line such as a telephone line, and dynamically holds the program for a short time. thing,
In this case, a program that holds a program for a certain period of time, such as a volatile memory in a computer system serving as a server or a client, is also included. Further, the above-mentioned program may be for realizing a part of the above-mentioned functions, or may be for realizing the above-mentioned functions in combination with a program already recorded in a computer system.

According to the present invention, by constructing the document search device and the document search system in a shared memory type, the broadcast performance is improved. In the case where search queries are simultaneously issued in a parallel computer, processing can be performed with short latency. Will be possible.
Further, by making the document search system a shared memory type, each of the document search devices (computers) that perform parallel processing can operate independently as agents, and in these parallel computers, It becomes possible to manage the computer efficiently.

In the document search system according to the present embodiment, by creating a document index in a stripe shape, each processor (document search device) performs a search process independently and can merge search results at high speed. . Also, by mirroring the document index, the search results can be fused at the highest speed using the search results that responded at the highest speed among the processors. Further, since the search results are stored in the shared memory in FIFO, the sort results of the new search results can be speeded up by using the search results.

[0087]

As described above, according to the first aspect of the present invention, a user query is input, and the input user query is morphologically analyzed using a morphological analysis dictionary. The user query processing of determining the words described in the above and determining the part of speech, inflected forms, etc. of each word, and recording the morphologically analyzed query in a shared memory space, and inputting a document to be processed, The input document is morphologically analyzed using a morphological analysis dictionary, the words of each sentence described in the document are determined, and the words and document names included in each sentence are stored in the document index database as an index of each document. The target document processing to be recorded and the presence / absence of the query stored in the shared memory space are constantly checked to confirm that the query has been recorded in the shared memory space. At the same time, the content of the query is read from the shared memory space, a document having a word group that matches the query is searched from the document index database, and when the document information that matches the user query is extracted, the shared A registration query process for recording the document information in the memory space, and constantly checking whether or not the document information recorded by the registration query process is recorded in the shared memory space, and as a search result for the user query, When such document information is described, it includes a search result process for reading out the document information, which is a search result recorded in the shared memory space, and outputting the document information to a user. Can be scalably processed.

According to the second aspect of the present invention, there is provided a document index database in which different document indexes are stored for each of a plurality of distributed document search devices, and the plurality of document search devices are provided in a shared memory space. A document search method in a document search system for performing a document search by sharing a document, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space. At the same time as confirming that the query was recorded in the memory space, the contents of the query were read from the shared memory space, and a document having a word group matching the query was searched from the document index database, and matched with the user query. Registration query processing unit that records the document information in the shared memory space when the extracted document information is extracted. And constantly check whether or not the document information recorded by the registration query processing is recorded in the shared memory space, and when appropriate document information is described as a search result for the user query, the shared memory And a search result process for reading out the document information, which is a search result recorded in the space, and outputting the document information to the user, so that each document search device constituting the document search system performs a search process independently, and Can be fused at high speed.

According to the third aspect of the present invention, there is provided a document index database in which the same document index is stored for each of a plurality of distributed document search devices, and the plurality of document search devices are provided in a shared memory. A document search method in a document search system that performs a document search by sharing a space, wherein each of the plurality of document search devices constantly checks whether there is a query stored in the shared memory space, At the same time as confirming that the query was recorded in the shared memory space, the contents of the query were read from the shared memory space, and a document having a word group matching the query was searched for from the document index database. When each of the document search devices records the search result in the shared memory space, the other document search devices record the search result. A registration query process for confirming that the other document search devices do not record search results during recording, and document information recorded by the registration query process in the shared memory space. Is constantly checked, and if appropriate document information is described as a search result for the user query, the relevant document information, which is the search result recorded in the shared memory space, is read and output to the user. Therefore, among the document search apparatuses constituting the document search system, the search result responding at the highest speed can be used, and the search result responding at the highest speed can be merged.

According to the fourth aspect of the present invention, a user query input unit for inputting a user's query and a user query input by the user query input unit are morphologically analyzed using a morphological analysis dictionary. A query morphological analysis unit that determines words described in the user query and also determines the part of speech, inflected form, etc. of each word, and a query that records a user query morphologically analyzed by the query morphological analysis unit in a shared memory space A user query processing unit having a recording unit, a target document input unit for inputting a document to be processed, and a morphological analysis of the document input by the processing target document input unit using a morphological analysis dictionary. Determine the words of each sentence described and record the words and document names contained in each sentence as a document index in the document index database. A target document processor and a document indexing unit, and the query confirmation unit for confirming the shared memory space constantly whether the stored query is in the,
A document index search unit that reads the contents of the query at the same time that the query confirmation unit confirms that the query has been recorded, and searches a document index database for a document having a word group that matches the query; When the index search unit searches and extracts document information matching the user query, a registered query processing unit having a search result recording unit that records the document information in the shared memory space; and A search result confirmation unit that constantly checks whether or not the document information recorded by the search result recording unit is recorded, and the search result confirmation unit includes:
When appropriate document information is described as a search result for the user query, a search result output unit that reads out the document information, which is the search result recorded in the shared memory space, and outputs it to the user With the processing unit, scalable processing can be performed for a large number of document indexes and a large number of user queries.

According to the fifth aspect of the present invention, there is provided a document search system in which a plurality of distributed document search devices perform a document search by sharing a shared memory space. Has a document index database in which different document index data is accumulated for each of the document search devices, and each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space. A query confirmation unit, and a document index retrieval unit that reads the contents of the query at the same time that the query confirmation unit confirms that the query has been recorded, and retrieves a document having a word group that matches the query from the document index database. Part, when the document index search unit searches and extracts document information that matches the user query, A registration query processing unit having a search result recording unit for recording the document information in the memory space; and constantly checking whether or not the document information recorded by the search result recording unit is recorded in the shared memory space. When appropriate document information is described as a search result for the user query, the search result check unit and the search result check unit read the document information, which is a search result recorded in the shared memory space, and And a search result processing unit having a search result output unit for outputting the search result, each of the document search apparatuses constituting the document search system can perform the search processing independently, and the search results can be merged at high speed.

According to the invention described in claim 6, there is provided a document search system in which a plurality of distributed document search devices perform a document search by sharing a shared memory space. Has a document index database in which the same document index data is accumulated for each of the document search devices, and each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space. And a document index that reads the contents of the query at the same time that the query confirmation unit confirms that the query has been recorded, and searches the document index database for a document having a word group that matches the query. When the search unit and each of the other distributed document search devices record a search result in the shared memory space, each of the other document search devices A search result recording unit that confirms that the document index search unit in the device does not record the search result, and prevents the document index search unit in each of the other document search devices from recording the search result during recording. A search result confirmation unit that constantly checks whether or not the document information recorded by the search result recording unit is recorded in the shared memory space; and the search result confirmation unit includes: A search result output unit that, when appropriate document information is described as a search result for the query, reads the document information that is the search result recorded in the shared memory space and outputs the read document information to the user Of the document search devices that make up the document search system, use the search result that responds the fastest, and respond the fastest. It can be fused to the search result that.

According to the present invention, a user's query is input, and the input user query is subjected to morphological analysis using a morphological analysis dictionary, and words described in the user query are input. User query processing to determine the part of speech, inflected form, etc. of each word, and record the morphologically analyzed query in a shared memory space; input a document to be processed; and perform morphological analysis on the input document. A morphological analysis using a dictionary to determine the words of each sentence described in the document, and a target document process of recording the words and document names included in each sentence as an index of each document in a document index database; It constantly checks whether there is a query stored in the shared memory space, and at the same time confirms that the query has been recorded in the shared memory space, When the content of the query is read from the shared memory space, a document having a word group that matches the query is searched from the document index database, and when the document information matching the user query is extracted, the document is stored in the shared memory space. Registration query processing for recording information, and constantly checking whether or not the document information recorded by the registration query processing is recorded in the shared memory space, and appropriate document information is described as a search result for the user query. And a search result process for reading out the document information, which is a search result recorded in the shared memory space, and outputting the document information to a user. Is recorded on a computer-readable recording medium, To read the recording medium on the recorded program in the computer system by executing, it can be processed scalable for a large number of document index and a large amount of user query.

According to the present invention, a document index database in which different document indexes are stored for each of a plurality of distributed document search devices is provided, and the plurality of document search devices are shared memory space. A document search method in a document search system for performing a document search by sharing a document, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space. Upon confirming that the query was recorded in the memory space, the content of the query was read from the shared memory space, and a document having a word group matching the query was searched from the document index database, and matched with the user query. Registration query processing for recording the document information in the shared memory space when the extracted document information is extracted, It is constantly checked whether or not the document information recorded by the registration query process is recorded in the shared memory space, and, as a search result for the user query, when appropriate document information is described, the shared memory space is A computer-readable program for causing a computer to execute a document search method in a document search system, comprising: a search result process that reads out the document information that is a recorded search result and outputs the document information to a user. Since the program is recorded on a simple recording medium, the program recorded on this recording medium is read into a computer system and executed, so that each of the document retrieval apparatuses constituting the document retrieval system performs a retrieval process independently and performs retrieval. The results can be fused quickly.

According to the ninth aspect of the present invention, there is provided a document index database in which the same document index is stored for each of a plurality of distributed document search devices, and the plurality of document search devices are shared memory. A document search method in a document search system that performs a document search by sharing a space, wherein in each of the plurality of document search devices, constantly checking whether there is a query stored in the shared memory space, At the same time as confirming that the query was recorded in the shared memory space, the contents of the query were read from the shared memory space, and a document having a word group matching the query was searched from the document index database. When each document search device records a search result in the shared memory space, each of the other document search devices records the search result. The registration query process for confirming that the search has not been performed, and for preventing the other document search devices from recording search results during recording, and the document information recorded by the registration query process in the shared memory space. It is constantly checked whether or not it is recorded, and when appropriate document information is described as a search result for the user query, the document information that is the search result recorded in the shared memory space is read out, And a computer-readable recording medium for causing a computer to execute a document search method in the document search system, which has a search result process for outputting the search result. The computer program is loaded into the computer system and executed. The in each document search apparatus constituting, utilizing the search results responsive to the fastest, can be fused to search results responsive to the fastest.

[Brief description of the drawings]

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

FIG. 2 is a flowchart showing processing contents of a query recording unit in a user query processing unit of the document search device shown in FIG.

FIG. 3 is a flowchart showing processing contents of a registration query processing unit A-3 of the document search device shown in FIG. 1;

FIG. 4 is a flowchart showing processing contents of a registration query processing unit A-3 of the document search device shown in FIG. 1;

FIG. 5 is a flowchart showing the contents of an inspection result processing unit A-4 of the document search device shown in FIG. 1;

FIG. 6 is a block diagram showing a configuration of an example in which the document search device according to the embodiment of the present invention is applied to a document search system on the Internet.

FIG. 7 is an explanatory diagram showing a configuration of a storage area provided for processing a user query in the shared memory of the document search system shown in FIG. 6;

8 is a flowchart showing specific processing contents of a query recording unit in the user query processing unit of the document search system shown in FIG.

FIG. 9 is a flowchart showing specific processing contents of a registration query processing unit of the document search system shown in FIG. 6;

FIG. 10 is a flowchart showing specific processing contents of a registration query processing unit of the document search system shown in FIG. 6;

FIG. 11 is a flowchart showing specific processing contents of a search result processing unit of the document search system shown in FIG. 6;

FIG. 12 is a flowchart showing an example of the details of the processing in step 30 ′ by the search result recording unit in FIG. 10;

FIG. 13 is a flowchart showing another example of the details of the processing of step 30 ′ by the search result recording unit in FIG. 10;

FIG. 14 is a flowchart showing the processing content of a flow A in FIG. 13;

[Explanation of symbols]

 A-1 User query processing unit A-2 Target document processing unit A-3 Registered query processing unit A-4 Search result processing unit D-1 Morphological analysis dictionary D-2 Processing target document database D-3 Document index database S- 1 Shared memory T-1 User query input section T-2 User query morphological analysis section T-3 Query recording section T-4 Target document input section T-5 Document index creation section T-6 Query confirmation section T-7 Document index search Part T-8 Search result recording part T-9 Search result confirmation part T-10 Search result output part

Claims (9)

[Claims] 1. A user query is input, and the input user query is morphologically analyzed using a morphological analysis dictionary to determine words described in the user query. User query processing for determining the shape and the like and recording the morphologically analyzed query in the shared memory space; inputting a document to be processed; morphologically analyzing the input document using a morphological analysis dictionary; The target document processing of determining the words of each sentence described in the above, and recording the words and the document names included in each sentence as an index of each document in a document index database, and a query stored in the shared memory space. Whether or not there is a query, and at the same time confirming that a query has been recorded in the shared memory space, Reading, searching a document having a word group that matches the query from the document index database, and extracting document information suitable for the user query, a registration query process of recording the document information in the shared memory space; and Constantly checking whether the document information recorded by the registration query process is recorded in the shared memory space,
When appropriate document information is described as a search result for the user query, a search result process for reading the document information, which is a search result recorded in the shared memory space, and outputting the read document information to a user. A document search method characterized by the following. 2. A document search system having a document index database in which different document indexes are stored for each of a plurality of distributed document search devices, wherein the plurality of document search devices share a shared memory space to perform a document search. A document search method in a system, comprising: in each of the plurality of document search devices, constantly checking whether or not there is a query stored in the shared memory space, and confirming that the query is recorded in the shared memory space. At the same time as the confirmation, the content of the query is read from the shared memory space, a document having a word group that matches the query is searched from the document index database, and when the document information matching the user query is extracted, the shared A registration query process for recording the document information in a memory space; and the registration query process in the shared memory space. Constantly check whether the document information recorded is recorded by Li treatment,
When appropriate document information is described as a search result for the user query, a search result process for reading out the document information, which is a search result recorded in the shared memory space, and outputting it to the user. A document search method in a document search system characterized by the following. 3. A document index database in which the same document index data is stored for each of a plurality of distributed document search devices, and the plurality of document search devices perform a document search by sharing a shared memory space. A document search method in a document search system, wherein each of the plurality of document search devices constantly checks whether there is a query stored in the shared memory space, and the query is recorded in the shared memory space. At the same time, the contents of the query are read from the shared memory space, and a document having a word group that matches the query is searched from the document index database. When recording in the shared memory space, confirm that each of the other document retrieval devices has not recorded the retrieval result, and record the document. A registration query process in which each of the other document search devices does not record a search result on the way, and constantly checks whether or not the document information recorded by the registration query process is recorded in the shared memory space, When appropriate document information is described as a search result for the user query, a search result process for reading out the document information, which is a search result recorded in the shared memory space, and outputting it to the user. A document search method in a document search system characterized by the following. 4. A user query input unit for inputting a user's query, and a morphological analysis of the user query input by the user query input unit using a morphological analysis dictionary, wherein words described in the user query are input. A query morphological analysis unit that determines the part of speech of each word, the inflected form, etc. A target document input unit for inputting a document to be processed, and a morphological analysis of the document input by the target document input unit using a morphological analysis dictionary to determine words of each sentence described in the document And a document index creation unit that records words and document names included in each sentence as a document index in a document index database. A target document processing unit, a query confirmation unit that constantly confirms whether or not there is a query stored in the shared memory space, and the query confirmation unit confirms that the query has been recorded, And a document index search unit for searching a document having a word group that matches the query from a document index database, and the document index search unit searches and extracts document information matching the user query. A registration query processing unit having a search result recording unit for recording the document information in the shared memory space; and constantly checking whether or not the document information recorded by the search result recording unit is recorded in the shared memory space. A search result confirmation unit, and the search result confirmation unit describes appropriate document information as a search result for the user query. If, share read the document information is a search result stored in the memory space, the document search apparatus characterized by having a retrieval result processing unit and a search result output unit for outputting to the user. 5. A document search system in which a plurality of distributed document search devices share a shared memory space to perform a document search, wherein the document search system has a different document index for each of the plurality of document search devices. Has a document index database stored, each of the plurality of document search devices, a query confirmation unit that constantly confirms whether there is a query stored in the shared memory space, and the query confirmation unit, At the same time as confirming that the query has been recorded, the content of the query is read, and a document index search unit that searches the document index database for a document having a word group that matches the query is searched by the document index database. If document information matching the user query is extracted, the document information is recorded in the shared memory space. A registration query processing unit having a search result recording unit; a search result confirmation unit that constantly checks whether or not the document information recorded by the search result recording unit is recorded in the shared memory space; and a search result confirmation unit. A search result output unit that reads out the relevant document information, which is the search result recorded in the shared memory space, when the appropriate document information is described as the search result with respect to the user query, and outputs the read document information to the user. A document retrieval system comprising: a search result processing unit having: 6. A document search system in which a plurality of distributed document search devices share a shared memory space to perform a document search, wherein the document search system includes the same document for each of the plurality of document search devices. A document index database in which an index is stored, each of the plurality of document search devices includes: a query confirmation unit that constantly confirms whether or not there is a query stored in the shared memory space; And a document index search unit that reads the contents of the query at the same time that the query is recorded and searches for a document having a word group that matches the query from the document index database, and other distributed documents. When the search device records the search result in the shared memory space, the document index search unit in each of the other document search devices performs the search. A registered query processing unit having a search result recording unit for confirming that the search result is not recorded, and a document index search unit in each of the other document search devices not recording a search result during recording; A search result checking unit that constantly checks whether or not the document information recorded by the search result recording unit is recorded in the shared memory space; and And a search result processing unit having a search result output unit that reads out the document information, which is a search result recorded in the shared memory space, and outputs the document information to the user when the document information is described. Document search system. 7. A user query is input, and the input user query is morphologically analyzed using a morphological analysis dictionary to determine words described in the user query. A user query process for determining the shape and the like and recording the morphologically analyzed query in a shared memory space; inputting a document to be processed; morphologically analyzing the input document using a morphological analysis dictionary; The target document processing of determining the word of each sentence described in the above, and recording the words and the document names included in each sentence as an index of each document in the document index database, and the query stored in the shared memory space Whether or not there is a query, and at the same time confirming that a query has been recorded in the shared memory space, Reading, searching a document having a word group that matches the query from the document index database, and extracting document information suitable for the user query, a registration query process of recording the document information in the shared memory space; and Constantly checking whether the document information recorded by the registration query process is recorded in the shared memory space,
When appropriate document information is described as a search result for the user query, a search result process for reading out the document information, which is a search result recorded in the shared memory space, and outputting it to the user. A computer-readable recording medium that stores a program for causing a computer to execute the document search method characterized by the following. 8. A document search apparatus having a document index database in which different document indexes are stored for each of a plurality of distributed document search apparatuses, wherein the plurality of document search apparatuses share a shared memory space to perform a document search. A document search method in a system, wherein in each of the plurality of document search devices, it is constantly checked whether or not there is a query stored in the shared memory space, and a query is recorded in the shared memory space. At the same time as the confirmation, the content of the query is read from the shared memory space, a document having a word group that matches the query is searched from the document index database, and when the document information matching the user query is extracted, the shared A registration query process for recording the document information in a memory space; and the registration query process in the shared memory space. Always check whether the document information recorded by Eri processing is recorded,
When appropriate document information is described as a search result for the user query, a search result process for reading out the document information, which is a search result recorded in the shared memory space, and outputting it to the user. A computer-readable recording medium on which a program for causing a computer to execute a document search method in a document search system is recorded. 9. A document which has a document index database in which the same document index is stored for each of a plurality of distributed document search devices, and wherein the plurality of document search devices perform a document search by sharing a shared memory space. A document search method in a search system, wherein each of the plurality of document search devices constantly checks whether or not there is a query stored in the shared memory space, and the query is recorded in the shared memory space. At the same time, the contents of the query are read from the shared memory space, a document having a word group that matches the query is searched for from the document index database, and each of the other distributed document search devices outputs the search result as the above. When recording in the shared memory space, confirm that each of the other document retrieval devices has not recorded the retrieval result, and record it. A registration query processing course by the other of each document search apparatus is prevented from recording a search result, the constantly check whether the document information recorded is recorded by the registration query processing in the shared memory space,
When appropriate document information is described as a search result for the user query, a search result process of reading the document information, which is a search result recorded in the shared memory space, and outputting the read document information to a user. A computer-readable recording medium on which a program for causing a computer to execute a document search method in a document search system is recorded.