JP2005018217A - Retrieval server device and retrival method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a distributed retrieval system for efficiently selecting the optimal node which stores retrieval object data to be retrieved in order to reduce traffic necessary for retrieval, and to reduce a communication load. <P>SOLUTION: This retrieval server device is provided with: a retrieval object database 361 for storing retrieval object data; a retrieval result history table 362 for storing selection data obtained by making a server identifier correspond to the evaluation data of another retrieval server device; a judging part 402 for retrieving the retrieval object database based on a retrieval request including a retrieval condition, and for transmitting the retrieval result and its own server identifier to a retrieval request origin when the retrieval object is detected, and for transmitting one part or whole part of the selection data associated with the retrieval request and the retrieval condition to another retrieval server device based on the selection data when the retrieval object is not detected; and a history updating part 408 for updating the selection data stored by its own device based on the selection data acquired from another retrieval server device. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a search server device and a search method for performing a search by search server devices distributed and connected to each other on a communication network.
[0002]
[Prior art]
Conventionally, a distributed search system in which a plurality of distributed search server devices are connected to each other via a communication line, and each search server device cooperates to perform a keyword search is known (for example, non-patent) Reference 1 and Non-Patent Document 2).
[0003]
In such a conventional distributed search system, when there is no data corresponding to the search server device requested to be searched, the search server device selects another search server device, and the selected search server device On the other hand, there is a method in which the process of sequentially transferring search requests is repeated until the corresponding data is found or another search server device is exhausted. Such a distributed search system will be described with reference to FIG.
[0004]
As shown in FIG. 10, each of nodes A, B, C, and D is a search server device, and each stores search target data representing a search target document and receives search request data including a search condition. The search is performed by comparing the keyword specified by the search request data with the character string constituting the document represented by the search target data.
[0005]
Further, in the above-described conventional method, the node A stores a search result history table storing search result history data representing the history of past search results. The search result history data is data generated for each search performed at each node, has items of keyword, node, and evaluation value, and represents at least one keyword used in the search; It consists of server data indicating the node and evaluation data indicating an evaluation value calculated based on the search result. The evaluation value increases as the number of hit search target data increases, and decreases as the number of hit search target data decreases, and is stored in association with each node.
[0006]
For example, when the node A receives search request data in which two search keywords “X” and “Y” are designated, the keyword using the keywords “X” and “Y” for the search target data stored by itself. Search is performed (process P1 in the figure). When the search target data is not hit in this search, the node A selects search result history data having data representing these search keywords as keyword items from the search result history table. At this time, if there are a plurality of corresponding search result history data, the search result history data having the maximum evaluation value indicated by the data of the evaluation value item is selected. Here, as shown in FIG. 10, since the evaluation value of the node D is the largest, the node D is selected.
[0007]
Next, the node A transmits search request data including two search keywords “X” and “Y” to the node D indicated by the node item data of the search result history data thus selected (processing in the figure). P2). The node D that has received the search request data performs a keyword search using the keywords “X” and “Y” for the stored search target data. When the search target data is hit in this search, the node D transmits data indicating the number of hit search target data and search result data including the search target data to the node A (process P3). The search result history table is updated based on the received search result data (process P4).
[0008]
As another distributed search system, as shown in FIG. 11, there is a search using metadata representing a character string and semantic information indicating the meaning of the character string. In this search, There is a method of generating search result history data based on information described in metadata and using it for selecting a search device.
[0009]
[Non-Patent Document 1]
NEUROGRID P2PSEARCH [Philosophy 21-08.01v03] Internet <URL: http: // www. neurogrid. net / php / phi1opy. php>
[0010]
[Non-Patent Document 2]
Neuro Grid P2P Search Distributed Search Internet <URL http: // www. jnutella. org / jnudev / jws # sam. lzh>
[0011]
[Problems to be solved by the invention]
In the conventional node selection method described above, the traffic required for the search is suppressed by selecting the transmission destination node of the search request data based on the information of the search result history table, and the purpose is to realize an efficient search. Yes.
[0012]
However, the above-described node selection method has a problem that the node cannot be selected when the keyword corresponding to the search request is not included in the search result history table. In addition, since there is a characteristic that the possibility of selecting an appropriate node increases as information in the search result history table increases in the past, a search result history table that enables effective node selection is created. A number of searches need to be done before it is done.
[0013]
Therefore, the present invention has been made in view of the above points, and in a distributed search system, it is necessary for a search by efficiently selecting an optimal node holding search target data to be searched. It is an object of the present invention to provide a search server device and a search method capable of reducing traffic and reducing a communication load.
[0014]
[Means for Solving the Problems]
In order to solve the above-described problems, the present invention provides a search server device distributed on a communication network and connected to each other, a search target database storing search target data to be searched, and other search servers. Search a search target database based on a search request including a selection data storage unit that stores selection data that associates a server identifier that specifies a search server device with evaluation data of the other search server device, and a search condition If the search target is detected, the search result and its own server identifier are transmitted to the search request source. If the search target is not detected, the search request is sent to another search server device based on the selected data. On the basis of the determination unit that transmits a part or all of the selection data related to the search request and the search condition, and the selection data acquired from another search server device, There characterized in that it comprises an updating unit that updates the selection data holding.
[0015]
In the invention described above, the format of the selection data may be that of which a search target is a character string and a metadata format including semantic information and a character string.
[0016]
According to the present invention as described above, each search server device receives a search request and, based on the evaluation data, transfers the search request to another search server device when the search server device does not hold the search target data. Since the search request transfer destination is selected, the optimum node can be efficiently selected and the traffic required for the search can be reduced. Further, in the present invention, when a search request is transferred, a part or all of the selection data held by the own device is transmitted to the transfer destination, so that the search processing efficiency in the transfer destination search server can be improved.
[0017]
In the above-mentioned invention, it is preferable that the updating unit has a function of acquiring selection data held by another search server device at an arbitrary timing and synchronizing it with the selection data held by itself. In this case, the selection data can be exchanged between the search server devices for synchronization, information can be shared, and a more optimal transfer destination can be selected. In the present invention, the selection data can be synchronized when the search node first starts the search, and at any time as required by the search node.
[0018]
In the said invention, it is preferable that the selection data memory | storage part has a function which stores the evaluation value acquired from the other search server, and the history of a server identifier as a history table. In this case, the selection data can be enhanced by accumulating the search results performed in the past as a history.
[0019]
The present invention relates to a search method in which search is performed by search server devices distributed on a communication network and connected to each other. The search target data to be searched is stored in a search target database. A step of storing selection data in which a server identifier for specifying the search server device and the evaluation data of the other search server device are associated with each other, searching a search target database based on a search request including a search condition, When the search target is detected, the search result is transmitted to the search request source, an evaluation value indicating the degree of matching with the search result, its own server identifier, and the search condition. When the search target is not detected, the selection data And a step of transferring a search request to another search server device.
[0020]
According to the present invention, each search server device receives a search request and transfers the search request to another search server device when the search server device does not hold the search target data. Since the transfer destination of the search request is selected based on the evaluation value of the search server device, it is possible to efficiently select the optimum node and reduce the traffic required for the search.
[0021]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0022]
(Constitution)
FIG. 1 is a block diagram illustrating a configuration of a distributed search system 1 according to the embodiment. As shown in the figure, the distributed search system 1 is a system that performs a search using nodes 30 (30A to 30G) that are distributed on the communication network 20 and connected to each other. The user terminal 10 is connected.
[0023]
The node 30 is a search server device, and realizes a function of performing a search process using a metadata database and a search result history table, which will be described later, in a general WWW (World Wide Web) server device. The user terminal 10 is a general computer device, starts a www browser, sends a request message in accordance with HTTP to the communication work 20, and receives a response message for the message.
[0024]
FIG. 2 is a block diagram illustrating a hardware configuration of each node 30. As shown in the figure, the node 30 includes a CPU 31 connected via a bus 33, a hard disk device 36, and a communication interface 32 that communicates with the communication network 20. The hard disk device 36 stores communication addresses of all nodes constituting the distributed search system 1, a search target database 361, a search result history table 362, and a search program 363, and the CPU 31 stores in the hard disk device 36. The stored search program 363 is read and executed.
[0025]
The search target database 361 stores a large number of search target data created using XML (extensible Markup Language). Each search target data has at least one metadata, and the metadata includes keyword data representing a character string (for example, “Shibuya”) as a search keyword and semantic information (for example, a character) indicating the meaning content of the character string. And semantic data representing the column “ground”).
[0026]
FIG. 3 is a block diagram showing functions constructed on the node 30 by the CPU 31 executing the search program 363. Here, a description will be given by taking as an example the node B to which the search request is transferred from the node 30A.
[0027]
As shown in the figure, the search request receiving unit 401, the determination unit 402, the search result transmission unit 403, the history reference unit 404, and the search request transmission are executed in the node 30B by executing the search program 363. A unit 405, a request content holding unit 406, a search result receiving unit 407, a history updating unit 408, a history table receiving unit 409, and a history table transmitting unit 410 are constructed.
[0028]
The search request receiving unit 401 is a module connected to the communication interface 32, and is addressed to the own node 30B transmitted from the other device (here, another node A) requesting the search via the communication network 20. Search request data is received, and the received search request data is input to the request content holding unit 406 and the determination unit 402. Here, the search request data includes search condition data representing a search condition and address data indicating a communication address of the requesting node A, and the search condition data is composed of at least one search condition. .
[0029]
The request content holding unit 406 is a storage device that stores the search request data until the search result data in response to the search request data input from the search request receiving unit 401 is returned. Is passed to the unit 408. The history table receiving unit 409 receives the evaluation table 61A transmitted together with the search request data from the search request data transmission source node 30A, and updates the information of the search result history table 362 via the history update unit 408. It is.
[0030]
The search result transmission unit 403 is a module that transmits search result data in accordance with an instruction from the determination unit 402. The search result data is composed of the communication address of the requesting device, the URL (Uniform Resource Locator) indicating the location of the extracted search target data, the number of extracted search target data, and the communication address of the own node 30.
[0031]
The determination unit 402 searches the search target database 361 based on the search request data including the search condition, and when the search target is detected, the search result for the node A, its own server identifier, and the search condition This is a module that transmits an evaluation value indicating the degree of coincidence with respect to, and forwards a search request to another search node based on a history table as selection data when a search target is not detected.
[0032]
The search processing in the determination unit 402 is realized by comparing the character strings while distinguishing the keyword data and the semantic data, and performing it between all the search conditions and all the search target data. When at least one search target data is extracted by this trial, the determination unit 402 instructs the search result transmission unit 403 to return the search result data indicating the search result to the device that has transmitted the search request data.
[0033]
In addition, the determination unit 402 requests the history reference unit 404 to select a node that is optimal for the search request represented by the search request data. Further, when a node is selected by the history reference unit 404, the determination unit 402 instructs the search request transmission unit 405 to transmit search request data to the node. However, when the communication address of the requesting device in the search request data is different from the communication address of any node 30, the determination unit 402 sets the communication address of the requesting device included in the search request data to the own node 30. After rewriting to the communication address, the above instruction is given.
[0034]
When a node is not selected, the determination unit 402 returns empty search result data including the communication address of the request source device and the communication address of the own node 30 to the source device of the search request data. The search result transmission unit 403 is instructed. In addition, the determination unit 402 acquires information necessary for creating the evaluation table 61A corresponding to the search node condition from the history reference unit 404, creates the evaluation table 61A, and transmits the evaluation table 61A to the history table transmission unit 410. Instruct. When the node is selected by the determination unit 402 described above, the history reference unit 404 refers to the search result history table 362.
[0035]
When requested by the determination unit 402, the history reference unit 404 represents the evaluation value represented by the evaluation value data associated with the metadata that matches the search metadata in the search request data by the node data. An evaluation index is calculated by adding each node, and a node having the largest calculated evaluation index is selected as an optimum node. However, if there is no metadata that matches the search metadata in the search request data in the search result history table 362, a node is selected according to a predetermined priority order. In addition, selection of a node having higher priority than the own node or the own node is prohibited.
[0036]
The search result receiving unit 407 receives the search result data for the search request data from the node selected by the history reference unit 404, and compares the communication address of the requesting device in the search result data with the communication address of the own node 30. To do. If the two do not match, the search result receiving unit 407 instructs the search result transmitting unit 403 to return the search result data to the device that has transmitted the search request data, and the request content holding unit 406 stores the stored search request. Discard the data. On the other hand, if the two match, the history update unit 408 creates search result history data based on the search result data and the search request data, and additionally stores the search result history data in the search result history table 362, and the search result data To the search result transmission unit 403, and the request content holding unit 406 discards the search request data.
[0037]
FIG. 4 is an explanatory diagram illustrating a specific data configuration example of the search result history table 362. As shown in the figure, the search result history table 362 stores search result history data created when a search result receiving unit 407 described later receives search result data from another node. In the present embodiment, the search result history data is associated with metadata constituting the search condition data and bibliographic data that bibliographically represents the result of performing the search using the search condition data at the other node. Data. This bibliographic data includes node data for identifying the node that created the search result data, semantic information indicating the semantic content of the search condition, a character string of the search condition, and the number of search target data hit in the search It is comprised from the evaluation value data showing.
[0038]
FIG. 5 is an explanatory diagram illustrating a specific data configuration example of the evaluation table. As shown in the figure, the evaluation data includes node data for identifying the node that created the search result data, semantic information indicating the semantic content of the search condition, a character string of the search condition, and a hit in the search And evaluation value data representing the number of search target data.
[0039]
(Operation)
Next, the operation of this embodiment will be described. Here, it is assumed that search result history data created based on past search results is stored in the search result history table 362 in advance.
[0040]
First, when the user operates the user terminal 10, two search conditions (here, metadata in which the character string “Shinjuku” is associated with the semantic information “place name” and the character string “French” have the semantic information “restaurant”. Assume that the associated metadata) is input. Thereby, the search request data including the search condition data composed of these metadata is transmitted to the node 30A via the communication network 20.
[0041]
FIG. 6 is a flowchart showing the procedure of the node selection process in the node A. As shown in the figure, the search request receiving unit 401 of the node 30A receives search request data (step S501). Then, the search request reception unit 401 inputs search request data to the determination unit 402.
[0042]
The determination unit 402 to which the search request data is input performs a search process for extracting search target data having search condition data from the search target database 361 (step S502). That is, the determination unit 402 performs processing for comparing character strings while distinguishing character strings and semantic information between all search metadata and all search target data. Here, if the information in the search result history table 362 is not sufficient and it is desired to acquire the search result history table 362 of another node (step S503: Yes), the search result history table 362 is acquired from the other node. (Step S504). As shown in FIG. 7, the retrieval result history table is acquired by transmitting a history table request to other nodes (nodes 30B and 30D in FIG. 7), and the retrieval result history table transmitted in response thereto. To get.
[0043]
On the other hand, if the search result history table 362 has already been acquired in step S503 in FIG. 6 (step S503: No), it is determined whether the evaluation table 61A has been received from the search condition transmission source (S505). When the evaluation table 61A is received from the search condition transmission source (step S505: Yes), as shown in FIG. 8, the search result history table 362 is updated based on the evaluation table 61A (step S505). S506). Here, since the node 30A is assumed, the evaluation table is not received.
[0044]
Using the search result history table 362 updated in this way, an optimum node corresponding to the search condition data is selected (S507). Thereafter, as shown in FIG. 9, the determination unit 402 creates an evaluation table 61A for the node to which the next search request is transferred (S508), and both the evaluation table 61A and the search request are sent to the selected node. Transmit (S509).
[0045]
Here, a method in which the history reference unit 404 selects an optimum node from the search result history table 362 will be described. The history reference unit 404 evaluates the metadata (“semantic information” = “place name”, “character string” = “Shinjuku”) and the metadata (“semantic information” = “restaurant”, “character string” = ”). The evaluation index is calculated by adding the evaluation value of “French”) for each node, and the node having the largest calculated evaluation index is selected as the optimum node.
[0046]
For example, in the node 30B shown in FIG. 4, the evaluation value of the search result history for (“semantic information” = “place name” “character string” = “Shinjuku”) is “60”. Further, since the history information (“semantic information” = “restaurant“ character string ”=“ French ”) is not stored, the evaluation index of the node 30B is 60 + 0 = 60. Similarly, for the node 30C, one piece of history data (“semantic information” = “place name”, “character string” = “Shinjuku”) is stored, and the evaluation value is “20”.
[0047]
One piece of history data (“semantic information” = “restaurant”, “character string” = “French”) is also stored, and the evaluation value is “20”. Therefore, the evaluation index of the node 30C is 20 + 20 = 40. For the node 30D, the evaluation value of the history data which is (“semantic information” = “place name”, “character string” = “Shinjuku”) is “20”, and (“semantic information” = “restaurant”, “character” Since the metadata “column” = “French”) is not stored, the evaluation index of the node 30D is 30 + 0 = 30.
[0048]
As described above, since the node 30B has the largest evaluation value, the history reference unit 404 selects the node 30B as the optimal node, and transmits information indicating that the node 30B is the optimal node to the search request transmission unit 405. . Based on the received information, the search request transmission unit 405 transmits search request data having search condition data and data indicating the communication address of the requesting node 30A to the node 30B (step S509 in FIG. 9).
[0049]
In this way, an appropriate transfer destination is selected at each node, the search request and evaluation table transfer, and the search result history table update are sequentially repeated to detect the requested search target and reply to the search request source. It is possible to efficiently select the optimal node and reduce the traffic required for the search. In particular, in this embodiment, when a search request is transferred, a part of the search result history table held by the own device is transmitted as evaluation data to the transfer destination, so that the search processing efficiency in the search server device at the transfer destination is improved. be able to.
[0050]
(Modification)
As mentioned above, although embodiment of this invention was described, this invention can be implemented with another various form, without deviating from the main characteristic. The above-described embodiments are merely examples of one aspect of the present invention, and the scope of the present invention is as shown in the claims, and modifications and changes belonging to the equivalent scope of the claims. Are all included within the scope of the present invention.
[0051]
As a modification, for example, in the above-described embodiment, the search result history table 362 is updated only when the search is started. However, the search result history table 362 is obtained from another node at an arbitrary timing. It is possible to update.
[0052]
【The invention's effect】
As described above, according to the present invention, in the distributed search system, by efficiently selecting the optimum node holding the search target data to be searched, the traffic required for the search is reduced, Communication load can be reduced.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a distributed search system according to an embodiment.
FIG. 2 is a block diagram showing a hardware configuration of a node according to the embodiment.
FIG. 3 is a block diagram illustrating functions that are constructed on a node when a CPU executes a search program in the embodiment;
FIG. 4 is an explanatory diagram illustrating a data structure of a search result history table according to the embodiment.
FIG. 5 is an explanatory diagram showing a data structure of an evaluation table transmitted together with a search request according to the embodiment.
FIG. 6 is a flowchart showing a procedure of node selection processing in the node according to the embodiment.
FIG. 7 is an explanatory diagram illustrating a history information acquisition operation in the embodiment.
FIG. 8 is an explanatory diagram illustrating an operation of history update processing in the embodiment.
FIG. 9 is an explanatory diagram showing operations of search information and evaluation table transfer in the embodiment.
FIG. 10 is an explanatory diagram showing operations of a conventional node selection process and history update process.
FIG. 11 is an explanatory diagram showing operations of a node selection process and a history update process using conventional metadata.
[Explanation of symbols]
1 ... Distributed search system
10: User terminal
20. Communication network
20. Communication work
30 (30A-30H) ... node
31 ... CPU
32 ... Communication interface
33 ... Bus
36. Hard disk device
61A ... Evaluation table
361 ... Search target database
362 ... Search result history table
363 ... Search program
401 ... Search request receiving unit
402: Judgment unit
403 ... Search result transmission unit
404 ... History reference part
405 ... Search request transmission unit
406 ... request content holding unit
407 ... Search result receiving unit
408 ... History update unit
409 ... History table receiving unit
410 ... History table transmission unit

Claims (4)

A search server device distributed on a communication network and connected to each other,
A search target database for storing search target data to be searched; and
A selection data storage unit that stores selection data in which a server identifier that identifies another search server device and an evaluation value of the other search server device are associated with each other;
The search target database is searched based on a search request including a search condition, and when the search target is detected, the search result and its own server identifier are transmitted to the search request source, and the search target is not detected. In this case, based on the selection data, a determination unit that transfers a part or all of the selection data related to a search request and a search condition to another search server device;
A search server device comprising: an update unit that updates selection data held by itself based on selection data acquired from another search server device. 2. The search according to claim 1, wherein the update unit has a function of acquiring the selection data held by the other search server device at an arbitrary timing and synchronizing the selection data with the selection data held by the update unit. Server device. The search server device according to claim 1, wherein the selection data storage unit has a function of storing an evaluation value acquired from another search server and a history of server identifiers as a search result history table. A search method in which search is performed by search server devices distributed on a communication network and connected to each other,
Search target data to be searched is stored in the search target database, and selection data in which a server identifier for specifying another search server device is associated with an evaluation value of the other search server device is stored. Steps,
The search target database is searched based on a search request including a search condition, and when the search target is detected, the search result and its own server identifier are transmitted to the search request source, and the search target is not detected. In some cases, the method includes a step of transmitting a part or all of selection data related to a search request and a search condition to another search server device based on the selection data.

Images