JP4018649B2 - Distributed search system, search device in distributed search system, information search method, and storage medium - Google Patents

Description

  The present invention seeks to find a computer having an information resource desired by a user in a computer environment in which a huge number of computers connected to a large-scale network each have some information resource and provide it to the user. Related to the function.

  Information and services provided by a computer are collectively referred to as “information resources” in the sense of intangible goods that are the basis of the information industry.

  As today's networks have evolved, the number of computers connected to the network has become enormous, and various services have been provided, what information resources each computer has? It has become very difficult to know.

  Even if such information is obtained, the network environment changes every moment due to computer or network maintenance or failure, so previously available information resources are not always available again. . Therefore, the user needs to know the computer that provides the desired information resource based on the latest information at the time of actually using the information resource.

  In general, there are many computers that have the same information resources, and each of them has the quality of information resources such as newness, accuracy, and abstraction level of each computer according to the management policy of the computer administrator. Of course it is different. Therefore, it is desirable for the user to find a computer having better information resources from among many computer groups. However, whether the information resource is good or bad is not understood unless it is actually used and other information resources are used and compared, and it takes time and effort for the user to do it, and it cannot be said that it is meaningful. It is also difficult for beginners with little knowledge of information resources to identify such things. Therefore, the function of recommending information resources recognized as relatively good by many users is also useful.

In recent years, many information resources on networks have been served by the World Wide Web (WWW). In the WWW, the location information of an information resource is expressed by a URL (Uniform Resource Locator). When a user tries to use an information resource, it is necessary to know the URL indicating the location information of the information resource. However, the URLs that individuals can find on their own are far less than all information resources on the network. Therefore, a search service called a search engine is provided on the WWW as a method for searching URLs corresponding to information resources. These methods can be basically divided into a step of collecting information on information resources available through the network and a step of managing the collected information and providing it to the user. Information collection methods can be broadly divided into the following two types.
Directory service The information resource provider asks the administrator (a kind of information resource broker) who provides the search engine service (directory service) to register in the directory list. /www.yahoo.com/) and AltaVista (http://altavista.digital.com/). In this method, the information resource provider makes a registration request with some confidence, so the quality of the information is good, but the registration work is often done manually by the administrator, and the burden on the administrator is heavy. Has disadvantages. As a result, information may not be updated promptly and accurately.
・ Robot Using a program called a robot, HTML (Hyper-Text Mark-up Language, a standard language for describing information provided on the WWW) that follows the links (anchors) in a document in order and exists all over the world. This is a method of automatically searching for and searching URLs and constructing a URL database. Examples include WWW Worm (University of Colorado, OAMcBryan) and RBS E Spider (University of Houston, D. Eichimann). The advantage of this method is that the information resource provider does not have to do anything. However, strictly speaking, it is impossible for a robot to discover unless it informs someone that it will provide information resources and links are provided. Further, since the information resource is referred to without the knowledge of the information resource provider, it is a problem that it is not known who should be notified that the information resource or service has been updated. Further, since the information resources are mechanically fished, it is easy to pick up useless information resources, and an unnecessary load on the network and the computer is generated.

Next, the method of managing the positional information related to the collected information resources and providing it to the user can be classified as follows.
・ Centralized management A system that provides all data on a single server, and many search engines, including Yahoo and AltaVista, are applicable. Although this method has an advantage that maintenance is easy because the management target is one place, on the other hand, since the access of the user is concentrated on one server, the load on the server tends to be very large. In addition, it is imperative that some users have a high communication cost to the server, and not all users can use it comfortably, and if one server goes down, it can no longer be used. It has the disadvantage of end.
・ Distributed management This is a method of managing and providing data by sharing it with several different servers, but it can be classified as follows according to the sharing method.

Load balancing is achieved by selecting and using a server that is easy for each user to access. This is the case with mirroring. The advantage of this method is that there are many servers with the same function, so that the service can continue even if a server goes down. However, this advantage cannot be received for the user unless the location information of the alternative server that performs the same service is known. Further, since all servers must always hold the same data, there is a problem that the cost required for data management is high.
Service distribution Classify services into several categories, and assign each category to several servers. This applies to DNS (Domain Name Service), which refers to the IP address from the computer name. A large-scale distributed database WAIS (Wide Area Information Server) can also be included in this category. In addition, there may be a case where access is distributed. In this method, maintenance is easy because the server to be managed differs depending on the type of service. However, if the type and range of services are limited, it can be regarded as centralized management after all, so the disadvantages of centralized management can be seen.

  Users must use different servers according to their desired services, and users feel inconvenient if there is no way to know the servers that provide them from the services (DNS uses the domain hierarchy). This is not a problem because WAIS does this automatically, and it does this with a database named directory-of-servers).

  On the other hand, the information resource recommendation function recommends what is preferable for users based on other people's recommendations and evaluation values, and other people's actions with the same preference, called social filtering or collaborative filtering. Technology has been developed. For example, Tapestry (Xerox Palo Alto Research Center, D.Goldberg, D.Terry) selects articles from a large number of articles that a person actively recommends, targeting online news and mailing list articles. It is a system that supports reading. On the other hand, the rating method for past articles is similar (similar in values). As a system for recommending items with high evaluation values by other users, the recommendation of net news is recommended. GroupLens (P.Resnick) to perform, Ringo (MIT, P.Maes, U.Shardanand) to recommend music albums.

  However, just because the preferences in one field are similar, there is no guarantee that the preferences in other fields are the same, so it may not always be good to follow the behavior and recommendations of a particular individual. is there. In addition, since information on preferences is centrally managed, the problems described in the centralized management of search engine information are highlighted regarding the management of preference data.

As described in the previous section, the problems in the existing methods can be summarized as follows.
-In directory services, registration is often done manually by the administrator, which places a heavy burden on the administrator. In addition, the search may not be performed well due to the mistake of the administrator.
・ Because robot programs do not carefully examine the contents of HTML documents, there is a high possibility of transferring useless HTML documents, which tends to increase the load on traffic and servers.
In order to keep traffic low, it is necessary to reduce the frequency of starting the robot program, and the collected information tends to be outdated, and the possibility that the acquired information is already invalid increases.
-Information gathering by robots may not be immediately reflected in search results even if the information resource provider does not know who should notify the change of the service content to be provided or even if it can be notified of the change There is sex.
If the database of information resources becomes huge, a huge amount of results are output for the search, making it difficult for the user to determine which is the most appropriate information resource. This is particularly noticeable when the user does not have sufficient knowledge about the target information.
-Just because a particular preference is similar cannot guarantee that all preferences are similar, so it is not always satisfactory with what a particular individual has recommended.
Furthermore, in the distributed search method as described above, how the agents are arranged and how they are combined has an influence on search performance. For this reason, the agent administrator needs to carefully consider how to construct the agent network in order to increase the overall efficiency.

  In the conventional distributed search method, the agent's neighbor relations are determined in advance by the agent administrator, and tuning must be performed manually by the administrator, so that the agent can flexibly cope with changes in the network environment of the agent. It was difficult.

  However, the network environment is constantly changing, and it is difficult to obtain search results promptly because the communication path between certain agents becomes unable to communicate due to a failure or the load on the agent temporarily decreases. There is a chance to go out into the situation. In addition, if the results from all the agents that have received the search request cannot be obtained, it is possible that a sufficiently desirable search result cannot be obtained.

  On the other hand, an agent that obtains resource information by returning search results may assume a situation in which useless resource information is meaninglessly stored and is not referred to by the nearest user. Such a situation is a situation that should be avoided in order to reduce the management cost of useless information.

  The present invention automates the management of information on information resources, mediates information resource advertisements and inquiries by users by information resource providers, and simultaneously advertises information resources when returning search results to users. Therefore, the object of the present invention is to find an information resource that solves the above problems by preventing the information from becoming obsolete and selecting one that is generally recognized as good from many information resources.

  Further, the present invention records information such as frequency, source, response time of other agents, etc., as advertisements of resource information received, search requests, or history of search results, and uses these data. The purpose is to efficiently propagate and search resource information by scheduling information transfer, allocating information delivery costs, and selecting an appropriate agent as a transfer destination.

  The present invention comprises a network of a plurality of agents, and FIG. 1 is a diagram illustrating the principle of each agent (search device) alone. In the figure, (1) is an advertisement processing unit that accepts advertisements related to information resources from information resource providers and notifies the information resource providers as necessary.The inquiry processing unit (2) This is the part that accepts search requests from users and returns the results to the users. The agent interface (3) is a part for exchanging advertisements and inquiries and the results with other agents. The information resource database (4) holds advertisements sent from the advertisement processing unit (1) and advertisements received from the agent interface (3).

  The information resource data control unit (5) stores the advertisement received from the advertisement processing unit (1) and agent interface (3) in the information resource database (4) and calculates the cost required for the advertisement. Correspondingly, the agent interface (3) is instructed to send it to other agents, and the information resource database (4) is used for inquiries from the inquiry processor (2) and agent interface (3) And the cost required for transmitting the search request is calculated, and the inquiry is forwarded to other agents through the agent interface (3). FIG. 2 schematically shows the exchange of data such as advertisements and inquiries in the entire system. In the figure, ○ represents an agent, and each of them performs the operation described here. Moreover, the solid line between agents in the figure represents the relationship that communication is possible.

  In the present invention, as shown in FIG. 2, a plurality of agents having the configuration described in the previous section constitute an interconnected infrastructure, and information on services that an information resource provider can provide in the form of advertisements. Notify the nearest agent. The agent who receives the advertisement stores the contents described in the advertisement in the information resource database, and performs the same advertisement to an agent within a range that can be transmitted within a certain cost. In this way, the agent who receives the advertisement keeps the advertisement for a certain period of time (referred to as the lifetime), erases it after that, and keeps old information from staying in the agent's information resource database indefinitely. Yes.

  On the other hand, when an inquiry from a user occurs, as with an advertisement, the inquiry is also sent to an agent within a range that can be transmitted within a certain cost, and when the inquiry reaches an agent having the corresponding data, the agent is sent. Returns the result to the user. When returning the results, the agents that the inquiry has passed through are traced in reverse, and each agent stores the results so that the advertisement is propagated outside the range advertised by the information resource provider first. (Fig. 3). When the user who has succeeded in the search uses the information resource and obtains a satisfactory result, the evaluation is fed back as shown in FIG. 4 and propagated in the same manner as the advertisement made by the information resource provider. An agent who has received a good evaluation about an information resource extends the lifetime of the information about the information resource so that the advertisement about the information resource that satisfies the user stays in the information resource database for a long time. Conversely, an agent that receives a negative rating shortens the lifetime of the information and makes it difficult to find by searching.

  Therefore, each agent only needs to communicate with neighboring agents, and further transmits only information specified by an advertisement or a search request, so that the communication cost can be kept low. In addition, advertisements on information resources that reflect users' usage performance in the form of survival time in the database, and are frequently used and gain popularity, exist in the database for a longer time than those that do not. Because it is propagated, it is easier to find when another user searches.

  Furthermore, as a modification of the present invention, a configuration in which a history database (5) is provided as shown in FIG. 23 in addition to the configuration described in FIG. This history database (5) is used to store the advertisement information, search requests, and search results transferred through the advertisement processing unit (1), inquiry processing unit (2), and agent interface (3). It is a database that records what content has been transferred and what it contains.

  The information resource data control unit (6) stores the advertisement received from the advertisement processing unit (1) and agent interface (3) in the information resource database (4), updates the history database (5), Depending on the contents of the message, the agent interface (3) is used to instruct processing such as sending to other agents. For inquiries from the inquiry processing unit (2) and agent interface (3), the information resource database (4) is searched, the history database (5) is updated, and the agent interface ( 3) Processes such as forwarding inquiries to other agents through Updating the resource information database (4) based on the search result obtained through the agent interface (3) is also performed by an instruction from the information resource data control unit (6). At this time, the history database (5) is updated in the same manner as in the case of the advertisement / search request.

  FIG. 24, FIG. 25, and FIG. 26 schematically show the exchange of such advertisement / search request / search result data in the entire system, respectively. The circles in the figure represent agents, each of which has the configuration of FIG. 23, and performs the above-described operation. Also, the solid line between agents in the figure indicates that the agents are in a neighboring relationship. The phrase “agents are in a neighbor relationship” here means that they are logically in a neighbor relationship, which means that agents can know each other's location information and communicate directly. Therefore, it does not necessarily reflect geographical or hardware (physical) proximity. Moreover, just because there is no neighborhood relationship does not mean that communication is not possible.

  In this configuration, as shown in FIG. 24, a plurality of agents having the configuration described in the previous section constitute an interconnected infrastructure, and information on services that can be provided by information resource providers in the form of advertisements. , Inform the nearest agent. Upon receiving the advertisement, the agent stores the contents described in the advertisement in the information resource database, and performs similar advertisements to neighboring agents. In this way, resource information is gradually transmitted to the agent.

  On the other hand, when an inquiry from the user occurs, the search request is transmitted to the neighboring agent as in the advertisement (FIG. 25), and when the inquiry reaches the agent having the corresponding data, Returns the result to the user. When returning the results, the agents that the inquiry has passed through are traced in reverse, and each agent stores the results so that the advertisement is propagated outside the range advertised by the information resource provider first. (FIG. 26).

  The agent of the information resource advertisement information, search request, and search result transfer destination is set by the agent administrator at the time of system introduction, but if the operation is continued to some extent, the advertisement information, search request, and search result History is accumulated. From this information, “the advertisement information coming from agent A is often related to the field F”, “search requests from agent B are often related to the field F”, or “search results from agent C are: The trend of resource information handled by agents such as “many things related to the field of F” can be seen.

  When a search request for information related to the field F is received from a certain agent using such information, the search request is preferentially sent to a neighboring agent close to the agent A or C, or directly the agent A or By sending it to C, it can be expected that the required information resource can be easily found.

  On the other hand, when an advertisement related to the field of F comes, if the advertisement information is transmitted to a person close to an agent with high demand for F, users in the area can easily find information about F. Can be expected. At the same time, with regard to the distribution of advertising information, it is necessary to immediately transmit what is in high demand, and to transmit what is not in the time when network traffic is low, thereby reducing network traffic concentration and agent overload. It is possible to prevent.

  In addition, if the information on the response time of the agent is recorded as history information, if there is a recent network outage or agent down situation, information transmission to the corresponding agent is temporarily performed. By stopping and sending to other agents, even if there is a failure in the agent network, the information search function can be provided without delay as a whole.

  According to the present invention, management of information on information resources can be automatically performed without relying on human resources, and the manager of resource information does not need to take time to register and maintain information. In addition, since the present invention is basically based on the declaration of the information resource provider as in the case of the conventional directory service, the quality of information that is an advantage of the directory service is not impaired.

  In addition, resource information is managed in a distributed manner, and by providing redundancy in the connection between agents, it is robust against changes in the network environment and improves search load distribution and ease of access for searchers. In addition, by limiting the cost that can be used for advertisement and search, it is possible to limit the range of agents that advertise resource information and the range of agents to be searched, so that the overall communication volume can be kept low. Further, since unnecessary access to the server of information resources such as robot programs is not required, a high load on the server is not generated. Furthermore, the information resource provider can always access the nearest agent and update the resource information. On the other hand, the information resource searcher receives feedback on the availability of the information resource, and other agents can Since the update information is transmitted immediately, it can be expected that the resource information is less likely to become obsolete.

  Feedback of information resource evaluation by information resource searchers adjusts the lifetime of resource information on each agent, so that information on useful information resources is easy to find, and resource information on other things is difficult to find Therefore, the resource information obtained as a search result is ranked based on the overall evaluation based on feedback from a large number of searchers. Therefore, the information resource search does not have the knowledge and skills to select many information resources. It can be expected that it will be easier for people to find, select and use better information resources. For this reason, the present invention greatly contributes to efficient distributed management and retrieval of large-scale resource information.

In addition, by adopting a configuration that maintains and uses the history of advertisement information, search requests and search results in a distributed search system for resource information, it is possible to distribute advertisement information in a time zone where the load of the agent is low, information to areas with high demand Distribution, preferential search to areas that have information, scheduling of advertisement distribution according to demand and network environment, routing of advertisement information and search requests by cost allocation according to past search results Therefore, it greatly contributes to traffic saving and efficient information distribution and information retrieval.

[First embodiment]
FIG. 5 is a block diagram of an embodiment of the present invention, showing an information resource discovery apparatus having information resource information distributed management and search functions. In the figure, the same numbers are given to those shown in FIG. 1 (numbers in FIGS. 11 and 13 indicate the order of data flow, and are not related thereto).

In normal operation, the information resource provider (producer) communicates information related to the service provided by the information resource provider (6) to the nearest agent in the form of advertisement in advance (advertisement request). As shown in Fig. 6, the specific format of the advertisement may be composed of the following items and the corresponding information (items marked with * are automatically processed by the advertisement processing unit (1)). (This information is not required by the information resource provider.)
ID (*): A symbol that identifies the advertisement.
From (*): Information that identifies the sender of the advertisement. For example, email address. Used for inquiries from users of information and as data for verification to prevent falsification of advertisement contents.
Subject: The name of the information resource to be provided.
Keywords: Keywords related to the information or services provided.
URL: Location information of the information to be provided. Since the position information is searched, such a field is provided, but it may be a field describing other information to be searched.
Maintainer: Information that identifies the administrator of the information resource. It can be the same as the sender of the advertisement.
Cost: A parameter that limits the advertising scope. The maximum number of agents that go through or the time spent on advertising. If a billing system is introduced for communication and computer usage, the amount may be used.
Date (*): Date and time when the agent received the advertisement.
Path (*): A string of IDs of agents to whom advertisements were relayed. The purpose is to prevent a single advertisement from being repeatedly transmitted to the same agent.

  The producer interface (6) transmits the advertisement to the advertisement processing unit (1) of the agent. Then, the contents of the advertisement are analyzed, the fields with the above * are supplemented, and the advertisement contents are passed to the information resource data control unit (5). Here, the lifetime of the advertisement is determined, and the contents of the advertisement are stored in the information resource database (4). The information resource data control unit (5) refers to the clock (8) and deletes the advertisement whose lifetime has already passed from the information resource database (4). The information resource data control unit (5) also calculates the cost required for the advertisement, and when the usable cost is still sufficient, the advertisement is transmitted to other agents through the agent interface (3).

Examples of the cost include the time from requesting an advertisement to the end of the advertisement, the communication cost, and the cost necessary for using the database.
In particular, by making the cost unlimited, it is possible to realize a conventional broadcast that does not limit the advertising range. This series of processing flow is shown in FIG.
In the figure, (6a) is the producer interface (6), (1a) is the advertisement processing unit (1), (5a) to (5c) are the information resource data control unit (5), (3c) is the agent interface (3) Correspond to each of the operations.

On the other hand, the information searcher (searcher) issues a search request to the inquiry processing unit (2) through the searcher interface (7) (search request). The specific format of the search request is similar to the format of the advertisement as shown in FIG. 8, and among the items, already known items are appropriately filled in, and items to be known by searching are made empty. (The handling of items with an asterisk (*) is handled automatically by the query processing unit (2), not by the searcher, as with advertisements). As a search performed by this, for example, the following search method can be considered.
1. Specify some keywords, leave the other items blank, and search for the names and URLs of information resources related to the specified keywords.
2. In order to check whether the service is actually performed or the contents of the service, specify the URL, leave other items blank, and search for Subject and Keywords.
3. In order to link people who provide the same resource, specify Subject and Keywords, leave the Maintainer, and search for people who manage related information resources. Conversely, specify Maintainer to search for information resources managed by that person.

Upon receipt of the search request, the query processing unit (2) sends a search request to the information resource data control unit (5), and the information resource data control unit (5) analyzes the request and based on the request, the information resource data control unit (5) Search the database (4). If a corresponding result is found, the result is returned to the search requester through the inquiry processing unit (2) and the searcher interface (7) (search result). At this time, as in the case of the advertisement, the information resource data control unit (5) calculates the cost, and if there is a surplus in the available cost, the agent interface (3) is used to search for another agent. Request, collect the results, and send them back to the information searcher. The search result is similar to the advertisement format, and supplements the corresponding data for the item to be searched in the search request. Regarding the cost, as mentioned in the information resource advertisement described above, time and cost can be considered. In addition, by making the cost unlimited, you can ask all agents to search without limiting the search scope, and conversely, by minimizing it, only one agent can be searched. it can. This series of processing flow is shown in FIG. In the figure, (7a) is a searcher interface (7), (2a) and (2b) are query processing units (2), (3b) is an agent interface (3), (5d) to (5i) are information resource data controls It corresponds to the operation of part (5).
[Second Embodiment]
FIG. 10 shows an embodiment for a large-scale distributed database. The user gives the keyword of the information to be searched to the database search front end (9), and the database search front end (9) first sends the location information of the database server having information on the keyword to the nearest agent to the searcher interface ( 7) Ask through. The agent that receives the search request executes the search according to the operation of the first embodiment described above. At that time, each agent compares the search results returned from other agents with the corresponding data in its own information resource database (4), and if the search results are newer, the information Claim 4 can be realized by rewriting the data in the resource database (4) with new data. Then, if the location information of the database server is obtained, the database search front end (9) selects some appropriate servers from the obtained results, transfers the keywords to those database servers, and stores the database. Perform a search for. Finally, search results from the database server are integrated and sent back to the user. FIG. 11 schematically illustrates this series of operations.
[Third embodiment]
FIG. 12 shows an example of another large-scale distributed database.

  The second embodiment is a method of executing a search after the location information of the database server is found by the present invention. This embodiment is a method for searching the location information of the database server and for the found database server. In this method, database searches are performed in parallel.

  The database server administrator advertises the database server in advance as in the previous section. A person who searches for information requests a database search through the searcher interface (7). The request is formatted into an appropriate format by the inquiry process (2) and passed to the information resource data control unit (5).

  The information resource data control unit (5) searches the information resource database (4) and searches for the location information of the database server that can respond to the search. If information about the corresponding database server can be found, the database server is caused to execute a search through the database interface (10). If there is a surplus in the usable cost, a search request is issued to other agents through the agent interface (3), and the search result is returned. On the other hand, if the result is returned from the database server that requested the previous search, the result is temporarily stored, and the system waits until the result from other agents is obtained. When all the search results have been returned, they are integrated and sent back to the search requester. FIG. 13 schematically illustrates this series of operations.

In this method, there is a possibility that the same search request is received from a plurality of agents to a certain database server, but an ID is assigned to the search request in advance, and the database server stores the search history together with the ID, and duplicate search is performed. It is desirable for the request to have a function of avoiding the same search many times, such as replying to the requester as “already answered”.
[Fourth embodiment]
In the above embodiment, the information on the information resource obtained from the agent is basically unilaterally given from the information resource provider side, but it is obtained as a result of the search as in the embodiment of FIG. The information resources may be actually used, and the information such as good or bad may be fed back to the agent. By doing so, it becomes possible for all users to use information resources effectively. For example, when an information resource is actually used, for an information resource that has already been found to be unusable, an advertisement is made so that the corresponding information resource is deleted from the information resource database of each agent. To prevent users from using the invalid information resource. In addition, if it is well suited to the user's wishes, advertisements for data control that extend the lifetime will be made so that other users can easily find information on the information resource. To do. This embodiment corresponds to claim 5.

  Feedback information related to the evaluation of the information resource is passed to the feedback processing unit (11) through the searcher interface (7). Then, it is converted into an appropriate format and passed to the information resource data control unit (5) for advertisement in order to be propagated to other agents. Upon receiving the advertisement, the information resource data control unit (5) updates the lifetime of the corresponding information resource entry in the information resource database (4) according to the evaluation, as long as there is a margin for the cost as in the case of a normal advertisement. The evaluation is transmitted to other agents through the agent interface (3). The processing flow of the feedback processing unit (11) is shown in FIG.

  In this figure, the evaluation from users is assumed to be three stages: + (satisfied),-(unsatisfied), and * (unusable). Then, for each evaluation, the information resource data control unit (5) is instructed to perform an advertisement requesting the extension, shortening, or deletion of the lifetime of the corresponding information resource entry in the database.

In the above example, the life of advertisements in the information resource database was adopted as a measure for measuring the quality of information resources. , Prioritize according to other evaluation criteria, sort, and return search results. In that case, when the information resource provider advertises and records information about the information resource in the information resource database (4) in each agent, the attribute of information resource evaluation is also recorded in the database. There is a need. When information about the same information resource is returned from a plurality of agents together with different evaluation values, the evaluation value for the entire agent organization is calculated by taking an average or the like. Finally, the searcher interface (7) does not present all search results to the information searcher, but may rank them according to the evaluation criteria of the searcher and present the best one.
[Fifth embodiment]
In the above embodiment, the evaluation value related to the information resource possessed by each agent is aggregated by searching by the information searcher and used by the searcher to select the information resource. The calculation is performed independently of the search, and may be used by the information resource provider to know the social evaluation of the information resource released by the information resource provider.

FIG. 16 is a configuration diagram of an information resource discovery method with a totaling function of information resource evaluation values by an information resource provider. The producer interface (6) accepts an evaluation request in addition to the advertisement request from the information resource provider. When the evaluation request is received, the information resource provider is checked if necessary, and the evaluation request processing unit Send the contents to (12). The evaluation request processing unit (12) analyzes the contents and sends an information resource data control unit (5) evaluation request. The information resource data control unit (5) searches for the evaluation value of the corresponding information resource in the information resource database (4), and sends it to other agents through the agent interface (3) in the same way as resource information advertisement and search. Also sends out an evaluation request. The evaluation values returned as a result of the search are collected in the information resource data control unit (5), compiled there, and returned to the information resource provider through the evaluation request processing unit (12) and the producer interface (6). .
[Sixth embodiment]
FIG. 18 shows the configuration of a distributed search engine that manages and searches information related to Web documents. In the figure, the same numbers are assigned to the same components as those of the above-described embodiments.

  This embodiment is technically similar to the first to sixth embodiments, but is disclosed as a best mode.

  In normal operation, the Web document creator communicates information about the Web document created by the Web document creator through the producer interface (6) to the nearest agent in the form of advertisements. The specific format of the advertisement is the same as in the above-described embodiments.

  The advertisement processing unit (1) assigns information such as registration date and time, information lifetime, initial value of advertisement cost / evaluation value based on the advertisement information about the Web document sent from the producer interface (6), Pass it to the Web document data controller (5). The Web document data control unit (5) is basically the same as the information resource data control unit, but the information resource targeted by this embodiment is a Web document, and is therefore named for convenience. The Web document database (4) is also an information resource database for the same reason. The Web document data control unit (5) registers the Web document information obtained by the advertisement in the Web document database (4), calculates the cost required for the advertisement, and reduces the cost, as in the above-described embodiments. If there is room, advertisements are transmitted to other agents through the agent interface (3). Also, based on the clock (8), Web document information whose validity period has expired is deleted from the Web document database (4).

  When receiving an advertisement request from another agent, the advertisement request is passed to the Web document data control unit (5) through the agent interface (7).

  The processing flow after the Web document data control unit (5) receives the advertisement request data from the advertisement processing unit (1) or the agent interface (7) is shown in FIG.

  On the other hand, a web document viewer searches for a web document through the searcher interface (7). The inquiry processing unit (2) creates a search request based on the input from the Web document viewer and sends it to the Web document data control unit (5). The format of the search request is as shown in FIG. The Web document data processing unit (5) searches the Web document database (4), searches for a Web document that satisfies the search request, and holds the search result. Then, as in the case of advertising, the remaining available cost is calculated, and if there is still room, the search request is forwarded to other agents through the agent interface (3), and the search result is returned. wait.

  When search results are returned from all agents that have issued search requests, or when a certain period of time has passed, the collected search results are sent to the inquiry processing unit (2), converted into a list format, and the searcher interface (7 ) To present to Web document viewers. When receiving a search request from another agent, the search request is passed to the Web document data control unit (5) through the agent interface (7), and the search result is sent to the requesting agent through the agent interface (7). Will be returned.

  FIG. 20 shows a processing flow after the Web document data processing unit (5) receives a search request from the inquiry processing unit (2) or the agent interface (7).

The agent that receives the search request directly from the web document viewer presents the search result to the web document viewer. Based on the search result obtained in this way, the web document viewer actually uses the searcher interface (7). Access the Web document shown in the search results. Then ask the viewer to evaluate whether the Web document meets the request of the viewer. The evaluation may be performed automatically by the searcher interface (7). Since it is not preferable that the actual Web document does not exist or cannot be used for some reason, it can be automatically collected as a negative evaluation. The processing flow of the searcher interface (7) at this time is shown in FIG.
[Seventh embodiment]
Web document collection can use not only advertisements from Web document creators but also Web document information of existing search engines. Agents can also be used as search engine wrappers to exchange information about Web documents with each other. In this way, existing search engines can be linked and integrated, and all search engines can be used in a unified manner.

  FIG. 22 is a configuration diagram of a distributed search engine using the existing search engine as described above. It is almost the same as the distributed search engine in the previous section, except that it has a search engine interface (13) and can obtain Web document information from an existing search engine.

  This search engine interface (13) extracts keywords included in the search request, performs a search using an existing search engine, and obtains the obtained Web document from the web document data control unit via the searcher interface. It is presented to the searcher.

  Web document database (4) caches Web document information of existing search engines and functions as a database of evaluation information related to Web documents. The Web document is analyzed, the advertisement information corresponding to the Web document is created in the advertisement information format as described above, and stored in the Web document database (4).

The basic operation for the advertisement request is the same as the previous section, but some operations for the search request can be considered according to the usage form of the existing search engine.
-A form in which the existing search engine is mainly used as a Web document database and the Web document database (4) is used as a database for storing evaluation information for Web documents.

In response to the search request, the search engine is searched through the search engine interface (13), the Web document database (4) is examined for each Web document in the obtained results, evaluation information is given, and the agent or Return it to the web document viewer.
・ The Web document database (4) is mainly used, but the existing search engine is used in an auxiliary manner when sufficient search results cannot be obtained.
In response to a search request, the Web document database (4) is searched first, and if a result suitable for the search request cannot be obtained or a sufficient number of Web document information cannot be obtained, the search engine interface (13) To search using existing search engines. The result is presented to the Web document viewer and registered in the Web document database (4) to prepare for the next search.
-A form in which an existing search engine is used to manage and maintain the document database (4).

  In response to a search request, search using the Web document database (4) as usual. At the same time, the search request history is stored, the search engine is searched through the search engine interface (13) based on the search request history at a certain time, and the Web document database ( Update 4).

  As described above, according to the present invention, management of information about information resources can be automatically performed without relying on human resources, and the manager of resource information does not need to take time to register and maintain information. In addition, since the present invention is basically based on the declaration of the information resource provider as in the case of the conventional directory service, the quality of information that is an advantage of the directory service is not impaired.

  In addition, resource information is managed in a distributed manner, and by providing redundancy in the connection between agents, it is robust against changes in the network environment and improves search load distribution and ease of access for searchers. In addition, by limiting the cost that can be used for advertisement and search, it is possible to limit the range of agents that advertise resource information and the range of agents to be searched, so that the overall communication volume can be kept low. Further, since unnecessary access to the server of information resources such as robot programs is not required, a high load on the server is not generated. Furthermore, the information resource provider can always access the nearest agent and update the resource information. On the other hand, the information resource searcher receives feedback on the availability of the information resource, and other agents can Since the update information is transmitted immediately, it can be expected that the resource information is less likely to become obsolete.

Feedback of information resource evaluation by information resource searchers adjusts the lifetime of resource information on each agent, so that information on useful information resources is easy to find, and resource information on other things is difficult to find Therefore, the resource information obtained as a search result is ranked based on the overall evaluation based on feedback from a large number of searchers. Therefore, the information resource search does not have the knowledge and skills to select many information resources. It can be expected that it will be easier for people to find, select and use better information resources. For this reason, the present invention greatly contributes to efficient distributed management and retrieval of large-scale resource information.
[Eighth embodiment]
FIG. 27 is a block diagram relating to the present embodiment, and shows an information resource discovery apparatus having information resource information distributed management and search functions. In the figure, the same numbers are given to those shown in FIG.

  First, the advertisement request process will be described with reference to the flowchart of FIG. In the normal operation, the information resource provider (producer) communicates information on the service it provides to the nearest agent through the producer interface (6) in the form of an advertisement (advertisement request). The specific content of the advertisement is the URL of the information resource to be provided, the name, a keyword representing the content, the administrator name, and the like. The producer interface (6) transmits the advertisement to the advertisement processing unit (1) of the agent. Then, analyze the content of the advertisement and check whether the necessary information has been dropped (for example, URL of information resources and related keywords), or whether invalid data is described (for example, whether the URL is in a valid format) Or the kanji code is normal) and the advertisement contents are passed to the information resource data control unit (5). Here, the history database (14) is updated based on the contents of the advertisement, and the contents of the advertisement are stored in the information resource database (4). The specific format of the history database (14) will be described later.

  Also, the advertisement information of the information resource can be transmitted to the agent determined in advance by the agent administrator through the agent interface (3). The agent that receives the advertisement information receives the advertisement information from the agent interface (3), passes it to the information resource data control unit (5), updates the history database (14), and updates the information resource database (4). Update.

  On the other hand, the information resource search request processing is sent from the information searcher to the inquiry processing section (2) through the searcher interface (7) as shown in the processing flowchart of FIG. Therefore, regarding the search request, the necessary data has not been dropped (for example, search keyword, email address if the result is returned by email), or invalid data is not included (for example, the email address is valid) Check whether the format is correct.)

  Upon receiving the search request, the query processing unit (2) sends a search request to the information resource data control unit (5), and the information resource data control unit (5) analyzes the request, and based on the request, the history database (14 ) Is updated. The information resource database (5) is searched, and the result is returned to the search requester through the inquiry processing unit (2) and the searcher interface (7).

  Also, in the case of search, as in the case of advertisement, it is also possible to request a search to a neighboring agent specified by the agent administrator through the agent interface (3), collect the results, and return it to the information searcher. it can. In this case, it continues to wait for a certain period of time until all the search results from the agent that has transferred the search request are obtained. When the search results are sent from the agent interface (3) to the information resource data control unit (5), the contents of the search results are analyzed (for example, whether the data has been corrupted during the communication process) and which search is performed. Check if it is a search result for the request. If the results for the search request have not yet been returned to the client (searcher or another agent), add the results to the return buffer that stores the search results to return to the client . Then, the search result is registered in the history database (5), and the contents of the information resource database (4) are updated. A series of processing flows for this search result is shown in FIG. In this way, the result is returned from all the agents that sent the search request or the clock (8) is referred to, and it waits until time-out, and the contents of the return buffer are returned to the client.

  As shown in FIG. 31, an example of a specific format of the history database includes a set of advertisement information, a time when the search request and the search result are received, their ID, a related keyword, and a sender.

  The ID is for identifying the received information, and has a format such as “R236666 @ A3”. The first character indicates advertisement information, search request, or search result, and is “P”, “Q”, and “R”, respectively. The number sequence that follows is a unique number sequence created from the time and process number to distinguish the information. After "@" is the agent ID.

  Next, the keyword column shows the related keyword of the information resource in the case of advertisement information, the search keyword in the case of a search request, and all the related keywords of the information resource (generally there are multiple) obtained as a search result in the case of a search result. It is.

The last sender agent indicates which agent has sent the information directly. With this database, it is possible to know from which agent information about which type of information resource is gathered or what type of information resource is in high demand.
[Ninth Embodiment]
FIG. 32 shows an embodiment of a distributed search apparatus having a spool of advertisement information. For users, search is more pleasing as soon as results are obtained, but advertisement processing is not necessarily so. It is preferable to postpone ad processing on the spot and not sacrifice performance of search processing just after accepting an ad request while performing search processing. is there. In fact, advertisements related to information resources that almost no one has searched for may be viewed by other agents immediately and may not be viewed by anyone. There may not be. On the other hand, if an information resource belongs to a field that has been searched many times by many users, the user will want to know immediately. is there. In the configuration of the present embodiment, for such a purpose, the advertisement information that does not need to be rushed for advertisement is temporarily stored in the advertisement information spool (10), and the advertisement processing is performed in batch.

  A series of processing flow in this embodiment is shown in FIG. The information resource data control unit (5) that has received the advertisement information searches the history database for a search request that uses a related keyword of the information resource included in the advertisement information, a synonym thereof, or a word related thereto. And a priority is determined from the result. For example, the number of search requests obtained by searching the history database (5) may be used. If the priority is greater than a certain threshold value, the advertisement processing is performed immediately, otherwise it is temporarily stored in the advertisement information spool (15). Then, the processing of the advertisement information in the advertisement information spool (15) is started when the agent is idle, when the advertisement information spool (15) is full, or at certain intervals. Also at this time, processing is performed in descending order of priority.

Since there is no history immediately after the activation of this apparatus, the priority is not particularly determined. Therefore, at first, the manager of the agent may decide the priority. After that, the priority determined by the administrator and the priority determined by the history may be used in combination.
[Tenth embodiment]
The purpose of this embodiment is to record route information in the history database (14) and perform more effective delivery of advertisements and search requests. The configuration diagram is the same as in FIG. 27, but the format of the history database (14) in FIG. 27 is the agent through which the information passes instead of the transfer source agent of the history database (14) as shown in FIG. The transfer route indicating is recorded. The information about the transfer route holds the ID of the agent that was passed in the exchanged information in order to prevent advertisement information, search requests and search results from being relayed repeatedly through the same agent (e-mail And the same as the NetNews Path field).

  The agent that has received the information records the route information in the history database (14). The keyword is a set of related keywords of the target information resource in the case of an advertisement and a search result, and is a set of keywords specified as a search word in the case of a search request. The transfer route is a string of agents through which advertisement information, search requests, or search results have been routed. For example, “A1: A2: A3” indicates that data has been transmitted from the agent A3 to the agent A1 via the agent A2.

  For example, in the example of FIG. 34, it can be seen that a large amount of advertisement information of the information resource related to the keyword kw1 is advertised by the agent A3 (in FIG. 34, the part marked with a white circle is applicable. The circles inside are for explanatory purposes and are not actually recorded in the database.) Resource information about kw1 is likely to be concentrated near A3. Therefore, when a search request for kw1 is received, it is easy to obtain many search results if it is directly transferred to the agent A3 that is not in the neighboring relationship.

Conversely, agent A7 finds that there are many search requests related to kw1 (the part marked with a black circle is applicable), and if the advertisement related to the information resource related to kw1 is directly transferred to A7, the search from A7 is effective. Can be done. Based on such considerations, the information resource data control unit (6) refers to the history database (5) as needed, determines an appropriate agent according to the contents of the advertisement information and the search request, and sends the advertisement information there. And transfer search requests.
[Eleventh embodiment]
In the distributed search apparatus employing the history database (FIG. 34) as in the tenth embodiment, an agent state table (16) is further provided as shown in FIG. The agent state table (16) is a table of response time required for transferring advertisement information, a search request and a search result to another agent. FIG. 36 shows an example of the table. Each column of the table consists of three items: the agent ID, the response time for that agent, and the last check time indicating the time when the response time was measured last. In the response time, “------” indicates that communication could not be performed normally.

  In this embodiment, when sending an advertisement request to a neighboring agent, if the advertisement request cannot be transferred to the neighboring agent due to a network failure or the like, the purpose is to send the advertising request to the nearest agent instead. It is said. It performs normal advertisement request processing and sends advertisement requests to nearby agents. Then, the response time at that time is measured, and the agent state table (16) is updated. If the advertisement request cannot be sent to the neighboring agent (the agent that could not be sent is A), the history database (14) as shown in FIG. Then, the agent state table (16) as shown in FIG. 36 is checked, the agent B having the fastest response is selected from the A neighbor agents, and the advertisement request is transferred there. By doing this, it is possible to avoid an agent that cannot perform processing, and to continue advertisement processing as a whole system. Also in the case of a search request, the same procedure can be used to avoid a failure and perform a search as a whole system.

  That is, as shown in FIG. 37, if an advertisement request, search request, or search result that has passed from Agent A via B in the past has been received, it can be known that B is in the vicinity of A, and A Even if a failure occurs, it can be sent directly to B.

FIG. 38 shows this series of processing flows.
[Twelfth embodiment]
In FIG. 39, when an advertisement request once placed in the spool is sent to a neighboring agent, the history database (14) is referred to, a time zone in which the network traffic and the agent load are relatively low is selected, and the transfer is performed. It is a block diagram of the distributed search apparatus which performs scheduling as it does.

  FIG. 40 is an example of an agent state table (16) for this embodiment. Unlike the example of the agent state table in the eleventh embodiment (FIG. 36), this table holds information on the average response time for each time zone and the number of times of actual measurement. These pieces of information are updated based on the time required for the search request and search result return processing that are mainly performed in real time. In this example, only the average response time and the number of measurements are recorded in the table. However, as with the table of the eleventh embodiment (FIG. 36), information regarding the final check time and the state at that time may be included.

In the case of receiving an advertisement request, as in the ninth embodiment, advertisement information with a low priority is temporarily stored in the advertisement information spool (15). At that time, the advertisement scheduler (17) creates an advertisement delivery schedule that defines which agent is to transmit the advertisement information in which time zone. For example, in FIG. 40, when the priority is very low, the advertisement information that must be sent to the agent A1 is transferred in the time zone from 01:00 to 01:59 where the processing of A1 is expected to be the least. If the priority is somewhat high, transfer is performed as early as possible even in a time zone where response time is somewhat long. Then, according to the advertisement delivery schedule, the advertisement information is delivered to the designated agent.
[Thirteenth embodiment]
FIG. 42 shows each neighboring agent according to past advertisements and search results recorded in the history database (14) in a distributed search search method in which the distribution range of information such as advertisement requests is limited using costs. It is a block diagram of the distributed search apparatus which has the function to distribute the cost spent on the transfer of the information sent to.

When an advertisement request is received, the normal advertisement request processing is performed, and when the advertisement request is transferred to another agent, the history database (14) is searched, and from which agent in the past, the advertisement is sent. Check if many matching search requests have been transferred. Agents who have received a lot of search requests can see that there is a lot of demand for that information, and if you spend a lot of cost in that direction and advertise, it's close to the high demand area. When the advertisement information is distributed until the next search request is generated, there is a high possibility that the resource information will be found at a low cost. It is the role of the cost manager (18) that distributes the cost of the advertisement request to the neighboring agents.

On the other hand, when a search request is transferred, when a search request is transferred to another agent, it is considered that there is a lot of corresponding resource information in the direction in which the advertisement request and the search result are transferred. If a search is performed with a lot of costs in that direction, it is possible to obtain a lot of search results and to find good resource information.
FIG. 42 is an example of cost distribution. In the past, if an ad request related to a keyword was received from neighboring agents A, B, and C at a ratio of 2: 5: 3, it would be obvious that there is a lot of information about that keyword in the direction of B. It is. Therefore, it is expected that better search results will be obtained if the cost of searching B is increased. Therefore, in this embodiment, the search cost is distributed at the ratio of 2: 5: 3, and the search request is transferred to each neighboring agent. Note that this ratio is not necessarily proportional, and may be configured to be weighted to some extent.

FIG. 2 is a diagram illustrating the principle of a search device (agent) in the distributed search system of the present invention. It is explanatory drawing regarding the advertisement and search of information resources in a distributed search system.

It is explanatory drawing of the propagation of the information regarding the information resource in a distributed search system. It is explanatory drawing regarding propagation of evaluation of the information resource in a distributed search system. It is a block diagram which shows the structure of the Example of a search device (agent). It is a figure which shows the example of an advertisement. It is a flowchart which shows the process of the advertisement in a search device (agent). It is a figure which shows the example of a search request. It is a flowchart which shows the process of the search in a search device (agent). It is a figure which shows the example of a search result. It is a block diagram which shows the structure of the search device in the large-scale distributed database of 2nd Embodiment. It is operation | movement explanatory drawing of the large-scale distributed database of 2nd Embodiment. It is a block diagram which shows the structure of the search device in the large-scale distributed database of 3rd Embodiment. It is operation | movement explanatory drawing of the large-scale distributed database of 3rd Embodiment. It is a block diagram which shows the structure of the search device in the large-scale distributed database of 4th Embodiment. It is a flowchart which shows the process of the feedback process part of FIG. It is a block diagram which shows the structure of the search device in the large-scale distributed database of 5th Embodiment. It is a block diagram which shows the structure of the search device in the large-scale distributed database of 6th Embodiment. It is a flowchart which shows the process of the advertisement request in 6th embodiment. It is a flowchart which shows the process of the search request in 6th embodiment. It is a flowchart which shows the process of the searcher interface in 6th embodiment. It is a block diagram which shows the structure of the search device in 7th embodiment. The principle block diagram of the search device which added the history database. It is explanatory drawing regarding the advertisement of the information resource in the distributed search system of FIG. It is explanatory drawing of the search of the information resource in the distributed search system of FIG. It is explanatory drawing of return of the information resource in the distributed search system of FIG. It is a block diagram which shows the structure of the search device in 8th embodiment. It is a flowchart which shows the process of the advertisement request in 8th embodiment. It is a flowchart which shows the process of the search request in 8th embodiment. It is a flowchart which shows the process of the search result in 8th embodiment. It is a figure which shows the example of the log | history database in 8th embodiment. It is a block diagram which shows the structure of the search device in 9th embodiment. It is a flowchart which shows the process of the advertisement request in 9th embodiment. It is a figure which shows the example of the log | history database in 10th Embodiment. It is a block diagram which shows the structure of the search device in 11th embodiment. It is a figure which shows the example of the agent state table in 11th Embodiment. It is explanatory drawing of the operation | movement in 11th embodiment. It is a flowchart which shows the process of the advertisement request by the search of an alternative agent in 11th embodiment. It is a block diagram which shows the structure of the search device in 12th embodiment. FIG. 29 shows an example of an agent state table in the twelfth embodiment. It is a block diagram which shows the structure of the search device in 13th embodiment. It is explanatory drawing of the operation | movement in 13th embodiment.

Explanation of symbols

1 advertising processing department,
2 Inquiry processing department,
3 Agent interface
4 Information resource database,
5 Information resource data control unit,
6 Producer interface,
7 Searcher interface,
8 clocks,
9 DB search front end,
10 DB interface,
11 Feedback processing unit,
12 Evaluation request processing unit.
13 Search engine interface
14 History database
15 Advertising information spool
16 Agent State Table 17 Advertising Scheduler
18 Cost manager

Claims (25)

A plurality of search devices are connected on the network, and each of the plurality of search devices stores means for storing advertisement information that is information including location information of information resources, a search request made to the search device, and the network. Accepts a search request sent from another connected search device, searches the advertisement information storage means of the search device itself, and sends the advertisement information to the search request source if the search target advertisement information is in the search device However, if the search target advertisement information is not in the search device, the search device is connected to each search device other than the search request source search device among other search devices that recognize the logical position information on the network. means for transferring the search request for the search result in which the advertisement information from other search device connected to the network, by following retrieval apparatus relays the search request to the contrary, While updating the advertisement information stored in each search apparatus on the transmission path, and means for returning to the requesting retrieval device, advertisements own search devices and other connected retrieval device to said network has A distributed search system for searching information storage means,
Each of the search devices determines whether to transfer the search request to another search device based on the cost information included in the information of the search request, thereby determining the information resource on the network according to the cost information. A distributed search system characterized in that a range of search devices for searching is determined. An evaluation value assigned to an information resource corresponding to advertisement information obtained by searching by an information searcher is stored in association with stored advertisement information on each search device on the network. Item 4. The distributed search system according to Item 1. The distributed search system according to claim 2 , wherein advertisement information having a good evaluation value and / or information resources corresponding to the advertisement information are selected from the search results and presented to an information searcher . Distributed search system according to claim 3, the information resource provider, and wherein the Rukoto inform the evaluation value of the information resources that they have advertised. Advertising information that is connected to the network and has means for storing advertising information that is information including at least the location information of the information resource, and that receives the search request from the user and retrieves the advertising information and extracts the advertising information; A search device for presenting information resources corresponding to the advertisement information,
An interface unit for accepting a search request from a search requester and presenting a search result;
A process of executing a search of the advertisement information storage means based on the received search request and transferring the search request to another search device connected to the network, and from another search device connected to the network Based on the transferred search request, a search of the advertisement information storage means is executed, and it is determined whether to transfer the search request to another search device based on the cost information included in the information of the search request. If the transfer request is determined, the search request is transferred to each search device other than the search request source search device among the other search devices that recognize the logical location information on the network. While updating the advertisement information stored in each search device on the return path while tracing back the search device that relayed the search request, the advertising information obtained by the processing and each processing described above, Retrieval system according distributed search system that wherein Rukoto and a control unit for executing the process of transferring the search device interface or transfer source is Yorimoto. 6. The control unit according to claim 5, wherein each of the control units calculates cost information included in the information of the search request, rewrites the cost information of the information of the search request, and transfers the information to another search device . Search device in a distributed search system. 6. The search device in a distributed search system according to claim 5 , further comprising a search unit that acquires information resources based on position information of information resources included in advertisement information of search results received by the interface unit . . The control unit acquires the information resource based on the location information of the information resource included in the advertisement information obtained by the search, and transfers the information resource to the request source interface unit or the request source search device. The search device in the distributed search system according to claim 5 . A feedback processing unit that receives an evaluation value of an information resource from an information searcher and determines whether to store advertisement information corresponding to the information resource and / or whether to delete the advertisement information based on the evaluation value As well as
The control unit is configured to change a period for storing the advertisement information stored in the advertisement information storage unit and / or delete the advertisement information based on a determination process of the feedback processing unit or a request from another search device. 6. The process according to claim 5 , wherein a process for performing a process for requesting execution of a process for changing a period for storing the advertisement information and / or a process for deleting the advertisement information to another search device is performed . Search device in a distributed search system. 6. The interface unit according to claim 5 , wherein advertisement information having a good evaluation value and / or information resources corresponding to the advertisement information are selected from the obtained search results and presented to the information searcher. Search device in distributed search system. A keyword included in the search request is extracted, a search process is performed using the keyword using an external search engine, and the obtained information resource data is sent to the interface unit for presentation to the information searcher Search device in a distributed retrieval system according to claim 5, characterized in Rukoto includes a search engine interface. 12. The distributed search system according to claim 11 , wherein the control unit analyzes the obtained information resource data to create advertisement information corresponding to the information resource, and registers the advertisement information in an advertisement information storage unit . Search device. A plurality of search devices each having an information resource and / or a database for storing location information for accessing the information resource are connected to the network, and each search device is based on a search request transferred from another search device. In addition to searching the database, the search request is forwarded to each search device other than the search request source search device among other search devices that recognize the logical location information on the network. A search method executed by the search device in a distributed search system configured as described above,
The database of the search device is searched based on a search request including search range information for setting a search range on the network transferred from another search device, and information resources and / or information resources that match the search conditions are searched. When the location information for access is extracted, each search on the return path is traced back to the search device that relayed the search request for the location information for accessing the information resource and / or the information resource. Updating the advertisement information stored in the device, and returning to the search device of the transfer source,
Determining whether to transfer the search request to another search device other than the transfer source of the search request based on search range information given to the search request transferred from the other search device; information retrieval method performed by the search apparatus characterized by chromatic and processing. The second process performs an operation on the search range information given to the search request, rewrites the search range information given to the search request to a value obtained by the calculation, and requests another search device to search information search method according to claim 13, characterized that you transfer. In the search range information, the number of search devices through which the search request passes on the network is set, and each search device subtracts the number of search devices included in the search range information given to the search request. information search method according to claim 1 4, characterized in that performing an operation The search request has unique identification information for each search request, and
A search history recording means for recording the search request for executing the first process and / or the second process together with the identification information;
When a search request is received from another search device, the search history recording means is searched using the identification information included in the received search request to determine whether there is a search request identical to the search request. Discriminate,
When the same search request as the search request does not exist, the first process and the second process are executed and the search request is recorded in the search history recording unit,
Information search according to claims 1 to 3, when said retrieval request the same search request and is present characterized in that it does not perform recording processing to the first processing and the second processing and the search history recording means Method. A plurality of search devices each having an information resource and / or a database for storing location information for accessing the information resource are connected to the network, and each search device is based on a search request transferred from another search device. In addition to searching the database, the search request is forwarded to each search device other than the search request source search device among other search devices that recognize the logical location information on the network. Computer-readable storage medium for storing a program executed by the search device in a distributed search system configured as described above
Because
The database of the search device is searched based on a search request including search range information for setting a search range on the network transferred from another search device, and information resources and / or information resources that match the search conditions are searched. When the location information for access is extracted, each search on the return path is traced back to the search device that relayed the search request for the location information for accessing the information resource and / or the information resource. Updating the advertisement information stored in the device, and returning to the search device of the transfer source,
Determining whether to transfer the search request to another search device other than the transfer source of the search request based on search range information given to the search request transferred from the other search device; A computer-readable storage medium storing a program for executing processing . The process of transferring a search request to another search device other than the transfer source is performed on another search device that recognizes logical location information on the network. The computer-readable storage medium according to claim 17 . The second process performs an operation on the search range information given to the search request, rewrites the search range information given to the search request to a value obtained by the calculation, and requests another search device to search the computer-readable storage medium of claim 1 7, characterized in that to transfer. In the search range information, the number of search devices through which the search request passes on the network is set, and each search device performs an operation of subtracting the number of search devices included in the cost information given to the search request. the computer-readable storage medium of claim 1 7, characterized in that to perform. The search request has unique identification information for each search request, and
Search history recording means for recording a search request for executing the first process and / or the second process together with the identification information in a recording device of a computer,
When a search request is received from another search device, the search history recording means is searched using the identification information included in the received search request to determine whether there is a search request identical to the search request. Discriminate,
When the same search request as the search request does not exist, the first process and the second process are executed, and the search request is recorded by the search history recording means, and the same search request as the search request exists. 19. The computer-readable storage medium according to claim 18 , wherein the first processing, the second processing, and the recording processing of the search history recording unit are not performed . The cost information is set with the number of search devices through which the search request passes on the network, and each search device performs an operation of subtracting the number of search devices included in the cost information given to the search request. distributed search system according to claim 1, characterized in that. And wherein the cost information are those number of search device the search request goes through over the network is set, the calculation of the cost information by the control unit is to subtract the number of the search device The search device in the distributed search system according to claim 6 . The search request has unique identification information for each search request, and
A search history recording means for recording a search request for executing a process of transferring a search request to the other search device together with the identification information;
When a search request is received from another search device, the search history recording means is searched using the identification information included in the received search request to determine whether there is a search request identical to the search request. Discriminate,
When a search request identical to the search request does not exist, a process for transferring the search request to the other search device is executed and the search request is recorded in the search history recording means, and the same search request as the search request is recorded. 6. The search device in the distributed search system according to claim 5 , wherein when there exists, processing for transferring a search request to the other search device and recording processing for the search history recording means are not performed . The process of transferring a search request to another search device connected to the network is performed on another search device that recognizes logical location information on the network. Item 6. The search device in the distributed search system according to Item 5 .

Images