JPH1091647A - Information accessing device and method capable of cooperatively utilizing knowledge related to information source - Google Patents

Abstract

PROBLEM TO BE SOLVED: To efficiently acquire required information from various distributed information sources while utilizing information related to information sources held by agents. SOLUTION: An agent part 13-1 generated based on an inquiry inputted by a network service space 11-1 accesses an object server 31 in a network service space 11-2 while moving on a wide area network 12. When sufficient information can not be obtained from a data base part 36, the agent part 13-1 acquires information related to an information source already accessed by another agent part 13-2 from a route information table 33 and accesses a server 38 in a service space 11-3 based on the acquired information. Thus knowledge is exchanged among plural agent parts through the table 33.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a distributed information system in which a plurality of information sources are distributed and arranged, and more particularly to an information access apparatus for searching required information such as objects in the system and a method therefor.

[0002]

2. Description of the Related Art Due to changes in the environment surrounding computers today, the computer environment using an information management system has changed from a centralized environment centered on a database server to a plurality of workstations connected by a high-speed network. Moving to a working distributed object management system.

[0003] Conventional database systems have primarily aimed at storing and providing numerical and character data. However, with the diversification of information, distributed object management systems must accumulate and provide not only simple data such as numerical values and characters, but also various types of objects such as procedures, programs, and data structures. ing.

[0004] These objects are distributed and stored in object servers connected to each other via a network, and a user provides an information providing service to an object server storing information required by the user. Request can satisfy the required requirements.

[0005] However, as the scale of the distributed information system of such a form increases, it becomes difficult for the user to know on which object server the required information is stored.

[0006] As one of the methods for solving this problem, there is an agent technology using an agent that works for a user. Here, agents are the functions that perform intellectual work on behalf of humans in the broadest sense,
In many cases, it is created as a process on a computer.

For example, in the case of an agent described in telescript, which is a language based on remote programming, the agent can autonomously move on a network to perform programmed processing (Japanese Patent Laid-Open No. 7-182).
174). Utilizing this property, there is a method of sending an artificial agent based on a user's request to an information source to obtain necessary information (Japanese Patent Laid-Open No. 6-301577).

There is also a method for enabling a dispatched agent to exchange information with another agent using a communication unit (Japanese Patent Laid-Open No. Hei 7-225798).
Furthermore, an agent that does not move but performs processing asynchronously while communicating with another agent is also conceivable.

[0009] According to such a technique, a user can obtain necessary information by dispatching an agent or causing communication between agents without knowing the location of an information source. Was.

[0010]

However, the above-described conventional agent technology has the following problems.

In the conventional agent technology, in order for an agent dispatched by a user to know a route for reaching an information source, it is necessary to obtain the route from a specific location server storing route information on a route to the information source in advance. Or, you have to explore it in a heuristic way.
For this reason, it takes a very long time for the user to obtain the answer to the inquiry from the agent, or it is difficult for the user to predict the time to obtain the answer.

In a large-scale distributed system including a plurality of information sources with autonomous updates, it is difficult to know which information sources have been updated and when. Therefore, when a location server for providing the route information to the agent is introduced, it is very difficult to collect the route information stored in the location server, and it is highly likely that the latest route information cannot be obtained. .

According to the present invention, necessary information is efficiently and quickly obtained from various information sources that are digitized and widely distributed while utilizing information on information sources such as route information held by an agent. It is an object of the present invention to provide an information access device and a method thereof capable of performing the above.

[0014]

FIG. 1 shows the principle of an information access apparatus according to the present invention. The information access device of the present invention includes three principles, a first principle, a second principle, and a third principle.

First, an information access apparatus based on the first and second principles corresponds to an information processing apparatus on a client side as an access source, and includes an input unit 1, an output unit 2, and an agent generation unit 3.

According to the first principle, the agent generating means 3 is at least one of a plurality of information sources, and at least one of the information sources accessed by a first agent unit which autonomously collects information. A second agent unit that acquires knowledge about an information source using the network 4 and acquires information from the at least one information source based on the knowledge is generated.

The output means 2 outputs information collected by the second agent unit. The agent unit used here is, for example, a process having a function of performing intellectual work on behalf of a human, and includes a movable agent described in a distributed processing description language and a non-movable agent fixed on a computer. Includes both agents. First
May be generated by the agent generating means 3 in the information access device, or may be generated by another information access device.

When the first and second agent units are movable on the network 4, the first agent unit moves to one of the information sources, accesses the information, and obtains knowledge about the information source. I do.

Here, the knowledge about the information source refers to, for example, identification information indicating the information source and usefulness information indicating the usefulness of the information source. The name or identifier of the information source is used as the identification information, and the score of the information source is used as the useful information. The first agent unit may move further while retaining the acquired knowledge about the information source,
The knowledge may be registered at an appropriate place on the network 4.

The second agent unit obtains knowledge about the information source obtained by the first agent unit by communicating with the first agent unit or by wandering over the network 4, and obtains the corresponding information source. Know the existence of. Then, after moving to the information source and acquiring the information, the information is returned to the information access device of the generation source, and the collected information is output from the output unit 2.

When these agent units cannot be moved, the first agent unit accesses one of the information sources by a method such as RPC (remote procedure call) and acquires knowledge about the information source. . The second agent unit communicates with the first agent unit to acquire knowledge about the information source acquired by the first agent unit and to know the existence of the corresponding information source. Then, the information source is accessed, and the obtained information is output from the output unit 2.

According to such a first principle, the second agent unit can know the existence of an information source without directly accessing it, and can collect necessary information therefrom. Therefore, the knowledge about the information sources among the plurality of agent units is effectively used. Further, by referring to the usefulness information for each information source, it is possible to efficiently find an unknown information source that has a high possibility of obtaining necessary information.

Further, the input means 1 inputs a request from the user to the agent generating means 3, and the agent generating means 3 generates the above-described second agent unit based on the request. As a result, the second agent unit efficiently collects the information requested by the user, and reports the obtained result to the user via the output unit 2.

According to the second principle, the agent generating means 3 autonomously accesses at least one of the plurality of information sources to acquire knowledge about the at least one information source, A second agent unit for transmitting the knowledge to the first agent unit using the network 4 is generated.

The output means 2 outputs information collected by the second agent unit. If the first and second agent parts are movable, the second agent part
Go to any source to access that information and gain knowledge about that source. Then, the knowledge is transmitted to the first agent unit by communicating with the first agent unit or registering the acquired information on the information source in an appropriate place on the network 4.

Thus, the first agent unit knows the existence of the corresponding information source and can access the information source. The second agent unit, after transmitting the knowledge about the information source to the first agent unit, returns to the information access device that has generated the information, and outputs the collected information to the output unit 2.
Output from

If these agent units cannot be moved, the second agent unit accesses any information source by a method such as RPC and acquires knowledge about the information source. Then, by communicating with the first agent unit, the knowledge is transmitted to the first agent unit, and the collected information is output from the output unit 2.

According to the second principle, similarly to the first principle, the first agent unit knows the existence of the information source without directly accessing and collects necessary information therefrom. Becomes possible. Therefore, the knowledge about the information sources among the plurality of agent units is effectively used.

Next, an information access device based on the third principle corresponds to a server-side information processing device existing in an information source, and includes an agent management means 5, an information storage means 6, and a knowledge storage means 7.

According to the third principle, the information storage means 6
Stores information. The knowledge storage means 7 is at least one of a plurality of information sources,
The knowledge of at least one information source accessed by the agent unit using the network 4 is stored.

Further, the agent management means 5 gives the information acquired from the information storage means 6 based on the access request of the second agent section to the second agent section, and sends the information to the plurality of information sources from the second agent section. And stores it in the knowledge storage means 7, extracts the knowledge about the at least one information source from the knowledge storage means 7, and provides it to the second agent unit.

When the first and second agent units are movable, the agent management means 5 firstly sends the first agent unit from the first agent unit arriving first from the access source.
Of the information source which the agent unit has already accessed, and stores it in the knowledge storage means 7.

Next, when the second agent unit arrives, it receives an access request from the second agent unit, accesses the information storage unit 6, and returns information conforming to the request. Further, the second agent unit receives the knowledge about the information source already accessed and stores it in the knowledge storage unit 7.

Further, the knowledge about the information source accessed by the first agent unit is extracted from the knowledge storage unit 7 and given to the second agent unit. Thus,
The second agent part knows the existence of the corresponding information source,
It will be possible to access that information source.

When these agent units cannot be moved, the first agent unit generated at the access source sends the knowledge about the information source that has already been accessed to the agent management unit 5 by a method such as RPC. The agent management unit 5 stores the knowledge received from the first agent unit in the knowledge storage unit 7.

Next, when an access request is sent from the second agent unit, the agent management means 5
It accesses the information storage means 6 and sends back information conforming to the request. Also, when the second agent unit receives the knowledge about the information source already accessed from the second agent unit, it stores it in the knowledge storage unit 7.

At this time, the agent management means 5 extracts knowledge about the information source accessed by the first agent unit from the knowledge storage unit 7 and sends it to the second agent unit. Thus, the second agent unit becomes aware of the existence of the corresponding information source and can access the information source.

According to the third principle, similarly to the first and second principles, the second agent unit knows the existence of the information source without directly accessing the information source, and extracts necessary information therefrom. It becomes possible to collect. Therefore,
Knowledge about each other's information sources is effectively used among a plurality of agent units.

According to the information access apparatus of the first, second, and third principles described above, when searching for a large number of information sources distributed on the network 4 using the agent, the information sources already accessed Information can be shared among a plurality of agents.

Therefore, each agent can select an appropriate information source to be accessed next from information sources already accessed by other agents, which is more efficient than searching for an information source at random. Information collection is realized. In addition, by sharing information indicating the usefulness of the information source that has already been accessed among a plurality of agents, information collection can be made more efficient.

For example, the input means 1 and the output means 2 shown in FIG. 1 correspond to the terminal device 21 shown in FIG.
, The agent generating means 3 corresponds to the application program 23 in FIG. 2, and the agent managing means 5 corresponds to the agent managing section 3.
Corresponds to 2. Further, for example, the information storage unit 6 corresponds to the database unit 36 in FIG. 2, and the knowledge storage unit 7 corresponds to the path information table 33.

[0042]

Embodiments of the present invention will be described below in detail with reference to the drawings. The information access device of the present invention can be implemented in the form of a client system or a server system connected to a communication network.
On the client, an agent unit is generated based on the user's request, and the agent unit moves across database servers on the network for information.

At this time, the agent unit sequentially writes the names of the information sources such as the servers visited from the client to the server in the route information table provided in the visited database server as the route information.
In addition, the utility of each information source is scored based on search conditions such as keywords passed from the user, and the search conditions and scores are added to the route information.

Next, the other agent unit refers to the route information left in the route information table, and if there is an unknown information source that has obtained a high score under the same search condition, the process moves to the database server. When such necessary operations are repeated and all necessary information is obtained, or when a predetermined time limit or the like is exceeded, the process returns to the client that generated the information and presents the information obtained at that time to the user.

As described above, by providing the route information table in which the route information is registered for each information source, it is possible for a large number of agent units to utilize each other's route information and search for the information source cooperatively. Become. Therefore, the usefulness of an unsearched information source can be predicted with a high probability, and information can be obtained more efficiently than when the information source is randomly moved.

The route information table of each information source includes:
Since the latest path information is sequentially written by the agent unit, the latest path information can be obtained without performing any special processing for collecting the path information. Therefore, the search for the information source is made more efficient.

FIG. 2 is a configuration diagram of a distributed information system provided with the information access device of the present invention. In FIG. 2, each network service space 11 operated autonomously
-1, 11-2, and 11-3 are mutually connected by a wide area network 12, and constitute a virtual-area wide area network service space.

Here, the network service space is
For example, it refers to a LAN (Local Area Network) in a company organization, and includes an information processing device such as a personal computer or a workstation connected to the LAN. This could also be a larger network integrating multiple LANs.

The network service space 11-1 represents a space on the client side that requests an information providing service, and the client 2 connected to the user terminal device 21.
2 inclusive. On the client 22, an application program unit 23 and an interface unit 24 are mounted as software.

The network service space 11-2
Represents a space on the server side that provides information, and includes an object server 31 that is a type of information source. The object server 31 includes a database unit 36 in which information is stored, and includes an agent management unit 32 and an interface unit 37 as software. Database unit 36
Stores various objects such as procedures, programs, and data structures in addition to data such as numerical values and characters.

The agent management unit 32 corresponds to, for example, a software engine including an interpreter for executing a procedure held by the agent unit 13-1.
A route information operation unit 34 for accessing the route information table 33 and a database operation unit 35 for accessing the database unit 36 are included.

The network service space 11-3
Represents another server space including another object server 38, and the wide area network 12 may be further connected to a large number of client-side network service spaces and server-side network service spaces.

Now, consider a case where a user belonging to the network service space 11-1 needs information stored in the database unit 36 of the object server 31 belonging to another network service space whose location is unknown. . The user inputs a request for information required by the user to the application program unit 23 of the client 22 via the user interface unit on the terminal device 21.

The application program unit 23 generates the agent unit 13-1 and transfers the input inquiry request to the agent unit 13-1 through the interface unit 24. The agent unit 13-1 is an artificial agent dispatched to an information source based on a user's request, and is a mobile process having a function of collecting information by connecting to the information source on behalf of the user. .

The agent unit 13-1 has a communication unit 14 for receiving a processing request from the interface unit 24, and holds the received processing request as inquiry information.

Agent section 13 storing inquiry information
-1 travels through the wide area network 12 while searching for a network service space to which the target information source belongs. At this time, the agent unit 13-1 holds, as the route information, the route followed when searching for the information source.

The agent unit 13-1 having reached the target information source establishes a connection between the interface unit 37 of the object server 31 and the communication unit 14 of the agent unit 13-1. Thereafter, the agent unit 13-1 sends the inquiry information to the database operation unit 35 of the object server 31. The database operation unit 35 performs an inquiry process with the database unit 36 based on the received inquiry information. By this inquiry processing, a search is performed under the specified conditions.

The agent unit 13-1 transmits the route information to the route information operating unit 34 through the interface unit 37.
And the route information operating unit 34 stores it in the route information table 33.

The result of the above inquiry process is returned to the agent unit 13-1 through the database operation unit 35 and the interface unit 37, and the agent unit 13-1
The result is held as answer information.

Agent section 13 storing answer information
1 satisfies the conditions requested by the user, returns to the user's network service space 11-1 via the wide area network 12, and returns the result to the application program unit 23 via the interface unit 24. The application program unit 23 performs necessary processing on the obtained result and outputs the result to the terminal device 21.

If the user's request cannot be satisfied, the agent unit 13-1 moves to another network service space 11-3 through the wide area network 12, and repeats the same processing. To determine the destination network service space 11-3, reference is made to the route information stored in the route information table 33 by the other agent unit 13-2, and an appropriate information source is selected from the information sources included in the route information. Select Then, the network service space 11-3 including it is set as the next destination.

The agent unit 13-1 repeats such movement and inquiry until a predetermined parameter exceeds a threshold. As this parameter, elapsed time, consumed cost, and the like are used.

The agent unit 13-1 that has not been able to satisfy the request even if the threshold value is exceeded is sent to the user's network service space 11 through the wide area network 12.
Return to -1 and reply to that effect.

FIG. 3 is a diagram showing a configuration example of an information processing apparatus (computer) corresponding to the client 22 or the object servers 31 and 38 in FIG. 3 includes a CPU (central processing unit) 41, a memory 4
2, a network connection device 43, an external storage device 45, and a medium drive device 46, which are connected to each other by a bus 44.

Each program corresponding to the application program unit 23, the interface unit 24, the route information operation unit 34, the database operation unit 35, and the interface unit 37 in FIG. 2 is stored on the memory 42 and executed by the CPU 41, for example. . As the memory 42, for example, a ROM
(Read only memory), RAM (random access memor)
y) etc. are used.

The interface units 24 and 37 are connected to the wide area network 1 through the network connection device 43.
2. Access a network such as a LAN connected to 2. The terminal device 21 in FIG. 2 is connected to the network, inputs an inquiry through the network connection device 43, and receives the result.

The external storage device 45 is, for example, a magnetic disk device, an optical disk device, a magneto-optical disk device or the like, and can store the above programs in addition to data.

The medium driving device 46 drives the portable recording medium 47 and accesses the stored contents. Portable recording medium 4
7 is a memory card, floppy disk, CD
Any computer readable recording medium such as a ROM (compact disk read only memory), an optical disk, a magneto-optical disk, etc. can be used. The portable recording medium 47 stores the above programs in addition to data.

Next, information held by the agent unit 13-1 will be described in detail. Agent unit 13-1
As shown in FIG. 4, can hold each information of inquiry information, answer information, route information, evaluation function, threshold value, and procedure. The contents of this information are as follows. (1) Inquiry information This is information on an inquiry when a user requests the agent unit 13-1 to collect information. In particular,
A search condition including a set of keywords representing necessary information and an arithmetic expression between them corresponds to this. For example,
The arithmetic expression “(AAA or BBB) and new product” represents information including the keyword “AAA” or “BBB” and the keyword “new product”. (2) Answer information Information obtained from an information source by an inquiry from the agent unit 13-1. (3) Route information This is information on each information source when one or more information sources are searched based on an inquiry. This information is configured by, for example, arranging the names and scores of the information sources accessed by the agent unit 13-1 in the order of access. By the route information processing described later, the route information of the agent unit 13-1 is written in the route information table 33 provided in each information source, and new route information is read from the route information table 33 into the agent unit 13-1. (4) Evaluation function Based on the score of the information source included in the obtained route information,
An algorithm for evaluating the usefulness of each source and determining the next source to visit. Each agent part 1
It is also possible to have a different evaluation function for each of 3-1 and 13-2. (5) Threshold This is a threshold for setting a search termination condition in order to prevent a situation in which information to be searched for when searching for an information source is not found and the search does not end forever. The agent unit 13-1 searching for the information source interrupts the search for the information source and generates the agent unit 13-1 when the parameter indicating the amount of resources consumed in the search process exceeds this threshold. It returns to the completed network service space 11-1.

As this parameter, an elapsed time, a search cost, and the like are used. The search cost is, for example, the number of accessed object servers. It is also possible to separately hold thresholds when one or more pieces of information are obtained and when no information is obtained. (6) Procedure This is a procedure program describing a series of processes performed by the agent unit 13-1 based on the inquiry request. According to this procedure, the agent unit 13-1 searches for an information source and moves in the network service spaces 11-2 and 11-3 while determining an information source to be visited by the evaluation function, and makes an inquiry to the information source. Run. Then, it returns to the original network service space 11-1, and returns the result of the inquiry to the user. Next, the contents of the route information stored in the route information table 33 will be described with reference to FIG. FIG. 5 shows an example of the route information table 33 in the information source Z.

In the route information table 33 of FIG.
“Keyword” means a phrase for determining required information based on an inquiry request. “Route” indicates an information source that has been traced by the agent unit that has registered the route information of the line, and “A (15) → B (40) →”
H (30)-> Z (80) ", the current information source Z
Indicates that the user visited another information source in the order of A, B, and H.

In this “route” item, the number written in parentheses after the name of each information source indicates the score given by the agent unit to each information source. The score is given to the result obtained from each information source using the keyword of the column included in the inquiry. In general, two agent parts that have exactly the same query information and evaluation function
Give the same score to the same source.

In FIG. 5, on December 11, 1995, the agent unit a563 having the same keyword “part PPP”
It can be seen that the agent unit a058 reaches the information source Z and gives a different score to the information source Z.
This is because the agent section a563 and the agent section a0
58 inquiry requests are, for example, “AAA parts PPP”
This is because they are different from each other, such as “parts PPP of BBB”.

Further, "agent" means an identifier attached to the agent section which registered the route information of the line, and "date" indicates a date when the agent visited the current information source Z. ing. The data in each row of the route information table 33 is added in the subsequent route information processing every time each agent unit performs an inquiry process.

Now, when the route information table 33 of the agent management unit 32 has the contents shown in FIG. 5, the agent unit 13-1 requiring the information on the keyword "salary" refers to the route information table 33. I do.
At this time, the agent unit 13-1 determines the next destination to be moved to the information source N visited by the other agent unit a015 before the current information source Z.

The reason for this is that the agent unit a01 which needed the same information on the keyword "salary" was needed.
5 is the record “J (2)-> B (23)-> N (95)-> stored in the route information table 33 on December 5, 1995.
According to Z (25) ", it is found that the agent unit a015 gives the highest score 95 to the information source N visited immediately before.

As described above, if the agent section which has required the information on the same keyword has already visited and there is an unknown information source among the relatively high-scoring information sources, the information source with the highest score is determined. An evaluation function can be set to be the next information source to visit. In addition to this evaluation function method, it is of course possible to use the route information of the route information table 33 to determine an information source to be moved next by another evaluation function.

For example, the agent unit 13-1 that needs the information about “AAA” records the information about “AAA” on December 5, 1995 and December 13, 1995 in the route information table 33 in FIG. From the agent section a7
Suppose that 52 has visited information source P twice.

At this time, the agent unit 13-1 compares the route recorded on December 5, 1995 containing the same information source P with the route recorded on December 13, 1995, We know that the score of P has increased from 50 to 89. From this, it is inferred that some update has occurred in the information source P from December 5 to 13, and the obtained information has increased.

Therefore, based on these route information, 9
Even if the information source P has already been visited before December 5, 2011, it is determined that updated new information can be obtained from the information source P, and P is determined as the next information source to be moved. Can be nominated.

As described above, if the date is described in the route information, the agent unit 13-1 knows that the information held by the information source has changed, and can perform a process according to the change.

As described above, various evaluation functions for determining the information source to be moved next based on the route information can be considered, and by providing the agent unit 13-1 with a function suitable for the request, It is possible to efficiently select an information source having necessary information.

The following methods can be considered as scoring methods for scoring information sources. For example, when searching for information including the keyword “HR”, if there is no information source that has information that matches the inquiry, the score of the information source for this keyword is set to 0.

If information conforming to the inquiry is found, the ratio of the total amount of information obtained from the information source to the total amount of information required by the agent unit 13-1 is calculated as a percentage. That is the score of the source. For example, if 100 pieces of information including the keyword “HR” are requested, if 50 pieces of information are obtained, the score will be 50 points.

Of course, other than this scoring method, any scoring method can be adopted, and the selection depends on the implementation of the information access apparatus. An example of the scoring method will be described in more detail later. Next, a series of processes performed by the agent unit 13-1 will be described in detail with reference to FIGS.

FIG. 6 is a flowchart showing the operation of the agent unit 13-1. When the process is started in FIG. 6, the user first inputs an inquiry request from the terminal device 21 (step S1). Next, the application program unit 23 of the client 22 generates the agent unit 13-1 (Step S2), and sends it to the information source (Step S3).

The agent unit 13-1 sent to the wide area network 12 searches for an information source based on the inquiry information (step S4), performs an inquiry process on the database unit 36 with the arrived information source, and as a result, Is received (step S5). Then, route information processing is performed (step S6), and it is determined whether or not the inquiry request is satisfied (step S7).

Here, if the request is not satisfied, it is determined whether or not a parameter used as a guideline for terminating the search exceeds a threshold value (step S8), and if not, the information source search is repeated (step S4).

If the request is satisfied in step S7, or if the parameter exceeds the threshold in step S8,
The agent unit 13-1 returns to the client 22 (step S9) and transfers the answer to the application program unit 23 (step S10). The application program unit 23 converts the received response into the terminal device 21 as a result of the inquiry.
(Step S11), and the process ends.

At step S 4, when determining the information source of the next destination to move to, the other agent unit 13-2 based on the route information in the route information table 33 so far.
Among the information sources included in the route information left by the etc., the information source considered to be the most suitable is selected.

As described above, by using the path information between the agents, it is possible to efficiently and quickly search for a necessary information source from a large number of object servers distributed over a wide area.

FIG. 7 is a flowchart of the generation process of the agent unit 13-1 in step S2 of FIG. When the processing is started in FIG. 7, the application program unit 23 first stores the agent unit 13-1 on the memory 42.
Is secured (step S21). Next, the procedure, the inquiry information, the evaluation function,
And threshold values are registered in order (steps S22, S23,
(S24, S25), the process ends.

Thus, a new agent unit 13-1 is generated in the execution environment of the client 22. When the returned agent unit 13-1 does not satisfy the inquiry request and generates an equivalent agent unit again,
Further, the route information held at the time of return is registered in the storage area.

Next, FIG. 8 is a flowchart of the information source search process in step S4 of FIG. The agent unit 13-1 normally holds the path information registered at the time of generation or the path information obtained from the path information table 33 of the information source, and performs a search process based on this.

When the process is started in FIG. 8, the agent unit 13-1 first reads the stored route information (step S31). Next, an evaluation function is applied to the obtained route information (step S32), and the information source with the highest score is determined (step S33). Here, it is determined whether or not the destination information source has been determined (step S34).
If it can be determined, move to the information source (step S3
5), end the processing.

If the information source has not been determined in step S34, it is next determined whether or not the above parameter has exceeded the threshold value (step S36). If not, the processing from step S31 is repeated. Also,
When the parameter exceeds the threshold value in step S36, the process returns to the client 22 (FIG. 6, step S9). The threshold value used for the determination in step S36 is the same as that in step S36 in FIG.
8 may be the same as or different from the threshold value used in FIG.

If the agent section 13-1 is a newly generated agent section and does not hold the path information, an appropriate information source is moved based on the designated information given from the user or the application program section 23. Choose first.

Here, an example of the algorithm of the evaluation function used in step S32 of FIG. 8 will be described. The path information held by the agent unit 13-1 describes a keyword that matches information currently required by the agent unit 13-1 and the score of an information source that provides the information.

When the agent unit 13-1 has, for example, the arithmetic expression “AAA and part PPP” as inquiry information, the route information read from the route information table 33 in FIG. 5 in step S31 is as shown in FIG. Become like In FIG. 9, from the path information table 33 of FIG.
It can be seen that the row of "PP" has been extracted.

In steps S32 and S33 of FIG. 8, information sources satisfying the respective conditions are obtained from the route information in the following order. [1] Known information sources that have already been visited are excluded from the information sources included in the route information. [2] Select an information source that includes both the keywords “AAA” and “parts PPP” and has the highest score. [3] Select an information source that includes one of the keywords “AAA” and “parts PPP” and has the highest score. [4] Among the information sources satisfying the conditions [2] and [3], an information source with the latest date is selected. [5] When there is no information on the information source satisfying the inquiry request in the route information, the conditions are satisfied in the order of [2], [3], and [4] including the information source excluded in the above [1]. Seek sources. [6] If the destination information source cannot be determined from the route information, the process returns to the client 22 (steps S34 and S34).
36, S9).

Assuming that the information source other than the information source Z in FIG. 9 is an unknown information source, the keyword “AAA” is obtained by the process [2].
The information source T (80) is selected as the information source including both the “company” and “parts PPP.” Next, in the process [3], one of the keywords “AAA” and “parts PPP” is included. Among the information sources F, P, S, and V, the information source P (89) having the highest score 89 is selected.

At this time, two recordings T (80) and P (8
Comparing the dates in 9), since P (89) is newer, the information source P is finally selected as the destination by the process [4].

Next, FIG. 10 is a flowchart of the moving process of the agent unit 13-1 in step S35 of FIG. The agent unit 13-1 is a program executed while moving on the network, and is an artificial agent of a user for collecting necessary information from an information source. The agent unit 13-1 has procedures and data therein, and moves along the wide area network 12 between the network service spaces 11-1, 11-2, and 11-3 based on the described procedures.

When the agent unit 13-1 moves, the procedure being executed is temporarily interrupted and frozen. When the movement is completed, the agent unit 13-1 that has reached the destination network service space is defrosted and resumes the processing that was being executed before being frozen.

Here, the freeze means that the agent unit 13-
The temporary variables that are the execution environment of the first procedure, the information in the memory 42 necessary for the execution of the procedure, the information about the procedure being processed, and the like are converted into a compressed data string for transfer through the network. Means to do.
The frozen agent unit 13-1 is transferred on the network using the same method as normal data transfer.

On the other hand, decompression means that the compressed data string is expanded again, the original procedure and the information in the memory 42 are restored, and the procedure executed by the agent unit 13-1 before freezing is restored. Restoring the execution environment.

The freeze / thaw processing can be incorporated as a part of the functions of the application program unit 23 and the agent management unit 32 in FIG. Here, a program for performing these processes is simply called a software engine.

When the processing is started in FIG. 10, the software engine of the information processing apparatus in which the agent unit 13-1 exists first saves the data in the memory 42 (step S41), and executes the currently executing procedure. Is saved (step S42). Next, the saved data is encoded (step S43), the procedure and the data are compressed (step S44), and transmitted over the network (step S45).

The software engine of the information processing apparatus to which the agent unit 13-1 has moved expands the received procedure and data (step S46), and restores the data in its memory 42 (step S47). Then, the procedure being executed before the movement is performed by the agent unit 13-1.
(Step S48), and the process ends.

Next, FIG. 11 is a flowchart of the route information processing in step S6 of FIG. In step S5 preceding the route information processing, the agent unit 13-1 that has reached the information source performs an inquiry process and obtains information to be sought from the information source.

When the processing is started in FIG. 11, the agent unit 13-1 first scores the result of the inquiry obtained from the information source (step S51),
The result is added to the route information held by the agent unit 13-1 and passed to the route information operation unit 34 through the interface unit 37 of the information source (step S52). The route information operation unit 34 registers the passed route information in the route information table 33 (Step S53).

Next, the agent unit 13-1 obtains new route information from the route information table 33 in the reverse procedure as information for determining the information source of the next destination (step S54). The process ends. Step S54
Then, only the route information including the keyword corresponding to the inquiry information of the agent unit 13-1 is acquired.

Incidentally, as a method of scoring the information source in step S51, the following example is given. (A) The information requested by the agent unit 13-1 is represented by a keyword K, and among all information amounts M of the information source R,
Let m be the amount of information that matches the keyword K. At this time, the score of the information source R is represented by m / M or (m / M) × 100 using percentage.

For example, the score given to the information source R by the agent unit 13-1 requesting the information represented by the keyword "AAA company" is 1000 for the total amount of information stored in the information source R, of which If 100 matches the keyword, their ratio was expressed as a percentage (100 /
1000) × 100 = 10 is the score of the information source R. (B) If the ratio of information matching the keyword K to other information sources is relatively small with respect to the score m / M of the information source R described in (a) above, the weight W1 is assigned to the score. Multiplication is also possible.

For example, when the score of the information source R is 10, it is assumed that it can be confirmed by some method that the same information does not exist in other information sources. At this time, the score is doubled to 2
It can be set to 0. (C) If all of the information requested by the agent unit 13-1 cannot be obtained from one information source, the information is obtained over a plurality of information sources. In this case, the weight W2 is added to the score of the information source R according to how much the request of the agent unit 13-1 is satisfied at each information source on the route where the agent unit 13-1 has searched for information. Is also possible.

For example, it is assumed that the agent unit 13-1 that requests information conforming to the keyword K satisfies the request by obtaining information across the information sources R1 and R2. In this case, if the score of R1 is 10 and the score of R2 is 30,
It is considered that information matching the keyword K exists relatively more in R2 than in R1. Therefore, 10 can be added to the score 30 of R2 alone to obtain 40. (D) When all the information requested by the agent unit 13-1 is obtained from one information source R, the score P of the information source can be multiplied by the weight W3 to obtain the score of the information source R. .

For example, when the score of the information source R is 5,
When the request of the agent unit 13-1 is completely satisfied only by the information source R, a score of 10 obtained by doubling the score 5 of the information source R can be used as the scoring result. (E) When there is very much information in the information source R,
If the ratio of the information that matches the keyword K is relatively small for all the information, the score of the information source R will be low in the scoring method (a). To prevent this, the scores need to be normalized.

Then, suppose that the total amount of information stored in another information source could be known by some method,
An average value v of the total amount of information of other known information sources is obtained in advance. At this time, by integrating the weight of the information source R with the weight M / v, the score can be normalized.

For example, if the total amount of information of the information source R is 100,000, the amount of information matching the keyword K is 100, and the average value of the total amount of information of known information sources is 1000,
The score of the information source R is (100/100000) × (10
(0000/1000) × 100 = 10. (F) As described in the above (a) to (e), besides determining the score based on the ratio of the number of information stored in each information source to the number of information matching the keyword, the information source R
May be used as the score of the information source R as it is.

When this method is used, it is possible to give a higher score to an information source having a larger number of stored information without using any knowledge about the keyword as to which information source the requested information exists in. become. This scoring method is based on the determination that the information source having a larger number of stored information items is more likely to have the desired information.

As described above, as a method of scoring information sources, first, each information source is considered to be independent.
There is a method in which the ratio of the amount of information meeting the request to the amount of information of each information source is scored. In addition, it is also possible to assign a more effective score by weighting, taking into account the relative high degree of satisfaction of an information source among a plurality of information sources.

In addition, when performing integration or addition by weighting, it is also effective to normalize scores so that scores can be assigned to all information sources on the same scale. Next, FIG.
An example of information source search in the wide area distributed information system of FIG. 2 will be described with reference to FIG. 12, the personnel information database 51, the AAA company database 52, the parts inventory database 53, and the terminal device 21 are connected by the wide area network 12.

Now, consider a case where a user requests information on a new product of AAA from the terminal device 21. First,
The agent unit 13-1 dispatched to the database 52 of the AAA company makes an inquiry to this database 52,
As a result, information belonging to the specifications of the new product is obtained.

However, since the details of the parts inventory are unknown, the agent management unit 3 in the network service space 11-2 to which the database 52 of the AAA company belongs is included.
2 is referred to. As a result, it is estimated that searching the parts inventory database 53 of the dealer is more promising than obtaining the personnel information database 51 in order to obtain information on parts inventory.

In this way, the agent section 13-1 knows the existence of the previously unknown parts inventory database 53, and can move there to obtain necessary information. In the embodiment described above, by adding a part of the registration information of the route information table 33, even if a failure or the like occurs in one of the information sources and it becomes impossible to access there, It becomes possible to detect in advance.

FIG. 13 shows an example in which such an extension is made to the route information table 33 of FIG. In FIG. 13, “failure information source” is used as the failure information item.
Has been added to the route information table 33. In this column, information on information sources that cannot be reached on the path from each agent unit to the current information source Z is described.

For example, "A->L" indicates that a failure or the like has occurred in the communication path from the information source A to the information source L, and the information source L cannot be reached. Further, “Q” indicates that a failure or the like has occurred in the information source Q itself and that access to the information source Q was impossible. here,
The value "-" indicates that there is no unreachable information source on the previous route.

Referring to FIG. 13, the agent unit a752
Record that the information source Q was unreachable on December 5, 1995. On the other hand, the agent unit a569 includes the information source Q on the path on December 10, 1995. From these data, the information source Q, which had been in a failed state on December 5, was restored later,
It can be inferred that it became accessible on day 0.

Therefore, the agent unit which referred to the route information table 33 from December 5 to December 10 can exclude the information source Q from the search target by regarding the information source Q as a failure state. Further, the agent unit which referred to the route information table 33 after December 10th,
The failure of the information source Q can be regarded as recovered and included in the search target.

As described above, by adding the fault information to the route information, when the agent unit 13-1 determines the information source to be moved next, the unreachable information source can be detected in advance. , And more efficient information collection is performed.

In the embodiment described above, an agent that can freely move on a network is used, but the present invention is not limited to this type of agent. The present invention is similarly applicable to any agent function that performs intellectual work on behalf of a human, and the agent may be a process fixed on a computer.

In the case of such an agent that does not move, by communicating with an agent on another computer,
Exchange information about the sources you have accessed so far,
The next information source to be accessed can be determined based on the obtained information.

[0133]

According to the present invention, when searching for a large number of information sources distributed on a communication network using an agent, access information on an already accessed information source is cooperatively used among a plurality of agents. It becomes possible.

By referring to the access information, each agent can select the next information source to be searched from among the information sources already accessed by other agents. Therefore, since the information sources to be accessed are narrowed down, it is possible to efficiently and quickly acquire necessary information from various information sources that are digitized and widely distributed.

[Brief description of the drawings]

FIG. 1 is a principle diagram of an information access device of the present invention.

FIG. 2 is a configuration diagram of a distributed information system including the information access device of the present invention.

FIG. 3 is a configuration diagram of an information processing apparatus.

FIG. 4 is a diagram showing information held by an agent unit.

FIG. 5 is a diagram showing an example of a route information table in an information source Z.

FIG. 6 is a flowchart illustrating an operation of an agent unit.

FIG. 7 is a flowchart of a generation process of an agent unit.

FIG. 8 is a flowchart of an information source search process.

FIG. 9 is a diagram illustrating an example of route information.

FIG. 10 is a flowchart of a movement process of an agent unit.

FIG. 11 is a flowchart of route information processing.

FIG. 12 is a diagram showing a search in the wide area distributed information system.

FIG. 13 is a diagram illustrating an example of a route information table including fault information.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Input means 2 Output means 3 Agent generation means 4 Network 5 Agent management means 6 Information storage means 7 Knowledge storage means 11-1, 11-2, 11-3 Network service space 12 Wide area network 13-1, 13-2 Agent section 14 communication unit 21 terminal device 22 client 23 application program unit 24, 37 interface unit 31, 38 object server 32 agent management unit 33 route information table 34 route information operation unit 35 database operation unit 36 database unit 41 CPU 42 memory 43 network connection device 44 Bus 45 External storage device 46 Medium drive device 47 Portable recording medium 51 Personnel information database 52 AAA company database 53 Parts inventory database

Claims (26)

[Claims] An information access device for a distributed information system having a plurality of information sources coupled by a communication network, wherein the information access device is at least one of the plurality of information sources and is autonomous. A second agent that acquires knowledge about the at least one information source accessed by the first agent unit that collects information using the network, and acquires information from the at least one information source based on the knowledge; An information access apparatus comprising: an agent generating unit that generates a unit; and an output unit that outputs information collected by the second agent unit. 2. The apparatus according to claim 1, further comprising an input unit for inputting a request from a user to said agent generating unit, wherein said agent generating unit generates said second agent unit based on said request. 2. The information access device according to 1. 3. The method according to claim 2, wherein the agent generation unit collects information while moving on the network, and generates, as the second agent unit, a process capable of acquiring the knowledge at a destination. The information access device according to claim 1, wherein 4. The process according to claim 2, wherein said agent generating means executes a process fixed on said information access device and capable of acquiring said knowledge via said network.
2. The information access device according to claim 1, wherein the information access device is generated as an agent unit. 5. An information access device for a distributed information system having a plurality of information sources coupled by a communication network, wherein the information access device autonomously accesses at least one of the plurality of information sources. Agent generating means for obtaining knowledge about the at least one information source and generating a second agent unit for transmitting the knowledge to the first agent unit using the network; and collecting the second agent unit. An information access device, comprising: an output unit that outputs the obtained information. 6. The method according to claim 1, wherein the agent generating unit is configured to acquire the knowledge including identification information indicating the at least one information source and usefulness information indicating the usefulness of the at least one information source. 6. The information access device according to claim 5, wherein the information access device generates two agent units. 7. The agent generating means, based on information collected from the at least one information source, scoring the usefulness of the at least one information source, and holding a scoring result as the usefulness information. Said second possible
7. An agent unit for generating a plurality of agent units.
Information access device as described. 8. The agent generating means may determine a ratio of the amount of the collected information to a total amount of information of the at least one information source, and use a score calculated from the ratio as the scoring result. 8. An information access apparatus according to claim 7, wherein said second agent unit is generated. 9. An information access apparatus for a distributed information system having a plurality of information sources connected by a communication network, comprising: an information storage means for storing information; and at least one of the plurality of information sources An information source, knowledge storage means for storing knowledge about the at least one information source accessed by the first agent unit using the network; and storing the information based on an access request from the second agent unit. Providing the information obtained from the means to the second agent unit, receiving knowledge about any of the plurality of information sources from the second agent unit and storing the knowledge in the knowledge storage unit, Agent management means for extracting knowledge about one information source and providing it to the second agent unit An information access device characterized by the following. 10. An information access apparatus according to claim 9, wherein said information storage means includes a database means for storing an arbitrary object. 11. The knowledge storage unit stores identification information indicating the at least one information source and usefulness information indicating the usefulness of the at least one information source as knowledge regarding the at least one information source. 10. The information access device according to claim 9, wherein: 12. The knowledge storage means stores path information on a path including the at least one information source, which is followed by the first agent unit on the network, and the agent management means stores the path information. The information access device according to claim 9, wherein the second agent unit is given to the second agent unit to notify the second agent unit of the existence of the at least one information source. 13. The knowledge storage means stores knowledge indicating that a change has occurred in the at least one information source,
The said agent management means gives the knowledge which shows that the said change occurred to the said 2nd agent part, and permits the said 2nd agent part to perform the process according to the said change, Claims characterized by the above-mentioned. Item 10. The information access device according to Item 9. 14. When the first agent unit detects a state in which access to the at least one information source is impossible, the knowledge storage unit stores failure information indicating the state, and 10. The information access apparatus according to claim 9, wherein the management unit gives the failure information to the second agent unit, and notifies the second agent unit of the state. 15. It has a plurality of information sources that manage arbitrary objects and operate autonomously on a communication network,
An information access device for a distributed information system in which information is collected by dispatching an artificial agent unit to an arbitrary information source, wherein: a means for connecting to the network; and The knowledge possessed by each agent unit, which is related to at least one of the plurality of information sources, is used in a cooperative manner, so that each agent unit can easily access any one of the plurality of information sources. An information access device, comprising: 16. An information access device for a distributed information system having a plurality of information sources coupled by a communication network, wherein the information access device is at least one of the plurality of information sources and is autonomous. Agent generation for acquiring knowledge about the at least one information source accessed by a first agent unit that collects information, and generating a second agent unit for acquiring information from the at least one information source based on the knowledge And an output unit for outputting information collected by the second agent unit. 17. An information access device for a distributed information system having a plurality of information sources connected by a communication network, wherein the information access device autonomously accesses at least one of the plurality of information sources. Agent generating means for obtaining knowledge about the at least one information source and generating a second agent unit for transmitting the knowledge to the first agent unit; and an output for outputting information collected by the second agent unit. And an information access device. 18. An information access device for a distributed information system having a plurality of information sources connected by a communication network, comprising: an information storage means for storing information; and at least one of the plurality of information sources. Information storage means for storing knowledge about the at least one information source accessed by the first agent unit; and information obtained from the information storage unit based on an access request from the second agent unit. The information is provided to the second agent unit, the knowledge about any of the plurality of information sources is received from the second agent unit and stored in the knowledge storage unit, and the knowledge about the at least one information source is stored from the knowledge storage unit. And an agent management means for taking out the information and giving it to the second agent unit. apparatus. 19. A recording medium recording a program for a distributed information system having a plurality of information sources connected by a communication network, wherein the recording medium is at least one of the plurality of information sources, Acquiring, using the network, knowledge about the at least one information source accessed by the first agent unit that autonomously collects information, and acquiring information from the at least one information source based on the knowledge; A computer-readable recording medium that records a program for causing a computer to realize a function of generating the second agent unit and a function of outputting information collected by the second agent unit. 20. A recording medium recording a program for a distributed information system having a plurality of information sources connected by a communication network, wherein the recording medium autonomously stores at least one of the plurality of information sources. A program for making a computer realize a function of accessing and acquiring knowledge about the at least one information source and transmitting the knowledge to a first agent unit using the network; A computer-readable recording medium that has been recorded. 21. A recording medium recording a program for a distributed information system having a plurality of information sources connected by a communication network, wherein the recording medium is at least one of the plurality of information sources, A function of storing, in a storage device, knowledge about the at least one information source accessed by the first agent unit using the network; and obtaining information obtained from a database based on an access request of the second agent unit. The second agent unit receives knowledge about any of the plurality of information sources from the second agent unit, stores the knowledge in the storage device, extracts the knowledge about the at least one information source from the storage device, and A program recording a program for causing a computer to realize the function to be provided to the second agent unit. Computer readable recording medium. 22. It has a plurality of information sources for managing arbitrary objects and operating autonomously on a communication network,
A recording medium recording a program for a distributed information system in which information is collected by dispatching an artificial agent unit to an arbitrary information source, wherein each agent unit has a plurality of agent units. A function of making use of the knowledge regarding at least one of the plurality of information sources in a cooperative manner and facilitating access to any of the plurality of information sources by each agent unit. A computer-readable recording medium on which a program for realizing the program is recorded. 23. An information access method for collecting information from a plurality of information sources coupled by a communication network, wherein the information access method is at least one of the plurality of information sources, A second agent unit for acquiring knowledge about the at least one information source accessed by the first agent unit performing the collection using the network, and acquiring information from the at least one information source based on the knowledge; And outputting the information collected by the second agent unit. 24. An information access method for collecting information from a plurality of information sources coupled by a communication network, the method comprising: autonomously accessing at least one of the plurality of information sources; An information access method, comprising: acquiring knowledge about the at least one information source; and generating a second agent unit that transmits the knowledge to a first agent unit using the network. 25. An information access method for collecting information from a plurality of information sources coupled by a communication network, wherein at least one of the plurality of information sources comprises a first agent Receiving, from the first agent unit, knowledge about the at least one information source accessed by the unit using the network; and acquiring information obtained from a database based on an access request of the second agent unit to the second agent unit. An information access method, wherein the information is provided to an agent unit, and knowledge about the at least one information source is provided to the second agent unit. 26. An information access method for managing an object and dispatching an artificial agent unit to a plurality of information sources operated autonomously on a communication network to collect information from the arbitrary information source Among the plurality of agent units, the knowledge possessed by each agent unit, and the knowledge regarding at least one information source among the plurality of information sources is cooperatively used, and each agent unit An information access method characterized by facilitating access to any of the plurality of information sources.