JP2001076002A - Information supply system provided with information needs estimation function - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance estimation ability for information selection reference and the selection method of a referring person by deciding estimation information selection reference based on data on the input of the referring person, the usable situation of display information and the selection of the information selection method itself, repeatedly obtaining and accumulating data on the referring person and correcting estimation information selection reference. SOLUTION: An agent 12 generates an electronic mail, based on the response of a referring person and delivers it to a mail server 21. The mail server 21 confirms the description of the agent name of the electric mail and transfers it to a mail generating program 22. The mail generating program 22 extracts data which is to be accumulated in an agent characteristic from the communication result of the electronic mail, calls the agent characteristic data from an agent DB server 23 and adds the extracted data to the agent characteristics. A characteristic decision subroutine is called from a subroutine group 25, and agent characteristic variables are updated/corrected.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a method for automatically collecting, selecting, and editing information from a large number of information on a network by an agent who has grasped a preference of a viewer who wants to browse information on a terminal device. The present invention relates to an information providing system.

[0002]

2. Description of the Related Art Recently, services for browsing and using Web sites by connecting to the Internet from portable telephones such as i-mode have become widespread. However, the display screen of the mobile phone is small, and the amount of information obtained at one time is small. Unless a pocket board or the like is used, it is troublesome to specify the desired information on the interface of the mobile phone, and in many cases, the information must be selected from limited options. This is the same situation in using the Internet on digital televisions, which are likely to become widespread in the future, since an interface called remote control is used. In addition, it is assumed that there are many vague situations that Internet users using such mobile phones and digital televisions (hereinafter simply referred to as viewers) do not specify the information they want and somehow want such information. Is done. In particular, when e-commerce on the Internet is assumed, companies try to promote products using media such as websites and e-mails. Estimate the products required by viewers and quickly display them on a small screen. It is very difficult. Assuming that consumers enter a store and select products in normal business transactions, consumers rarely specify products from the beginning, and it is usually difficult to identify which product genre is desired. Frequently, the user is wondering what to do, or when they buy the product they want, they buy the product they see. This situation itself is the pleasure of shopping. Entering the store because it is somehow fun,
There are even situations where they bought a lot when they came out. Since electronic commerce is also a form of commerce, it is a matter of how quickly and accurately to display products that match potential needs. However, in the current website information service, some sites such as Amazon.com, which have a function of displaying a homepage customized for each visitor, have achieved some results. There is no system that grasps and estimates potential needs, provides information and product information in a smart way, and further refines the way to search for that information. Even if you try to get a deeper information of a reader by answering a huge number of items and getting answers to find the viewer's wants, if it is troublesome, the response rate will be low and the wandering viewer who changes every day It is difficult to keep track of wants.

[0003] In order to meet the needs of viewers on such small display screens and remote-controlled terminal devices, to accurately display information in accordance with wants, to quickly display information, and to provide nimble information to viewers whose purpose is ambiguous. With mail software with characters limited to the function of carrying mail such as post pets, the only way was for the information provider to guess needs and select information. A system that customizes homepages such as Amazon.com and Isize and distributes them to users is a system in which a specific business operator customizes based on customer information obtained by the company and the information is provided by the viewer itself. It is not an information providing system that does not specify the information, or selects the method of estimating the method and the method itself. Further, it is not created by adding the result of inputting the preference of the viewer himself while protecting the privacy of the individual. Since the system is not owned by the viewer, it is not possible to copy the agent to other viewers at the viewer's will. Neither the character is set based on the selection of the viewer, nor is the information amount adjusted to suit the day of the viewer. Viewers' reaction to a somewhat random distribution of information does not have the effect of finding unexpected potential needs. Unless the viewer has a clear purpose for collecting information, the viewer will not provide detailed information selection conditions for a specific company, and if he is not aware of that purpose, the viewer will be Will not try to get. In an existing system such as an agent that collects information autonomously by setting certain conditions, processes such as learning and growth of the agent are missing, and its use value is naturally limited.

[Problems to be solved by the invention]

[0004] Therefore, in the conventional method, a viewer's personal agent that naturally grows while the viewer uses it on a daily basis searches for conscious needs and subconscious wants of the viewer, and obtains various information. I couldn't expect to select information from the information in a sensible way for the viewers themselves, or even to provide them while protecting the privacy of the viewers.

[Means for Solving the Problems]

[0005] Therefore, an individual agent of the viewer is resident at the terminal of the viewer, and the agent of the viewer derives his / her preference by talking with the viewer every day. , Based on purchase histories, survey responses, etc., using data mining or other methods to estimate and provide products, services, and other information required by viewers, Even when browsing on a small display screen such as a mobile phone, it is possible to provide information that matches the wants of the viewer with higher accuracy. As a result, it is a powerful weapon for businesses to promote e-commerce.

What is called an agent here is:
It is a collection of programs and data that operate autonomously on behalf of the viewer, and may be located on the viewer's terminal (client) or a server on the network. May be moved. An agent may have a program or data independent of each other, or a program or data of an individual viewer and a program or data common to the system may be operated in conjunction with each other.

[0007]

BEST MODE FOR CARRYING OUT THE INVENTION An agent set in a terminal of a viewer edits the information collected and selected by appropriately adding an index, a greeting, and other information, and distributes the information to the terminal of the viewer.
indicate.

[0008] The agent takes in the viewer's input to the transmitted information, the opening status of each piece of information, the browsing time, the order information of the product, etc. as the original data for setting and correcting the characteristics of the agent, and the agent up to that time. Modify the traits yourself. If the characteristics of the agent are defined by the original data and a plurality of variables (attribute name, attribute value, etc.), a calculation formula for modifying the variables representing the characteristics that have been obtained is determined in advance, and For example, if "movie" information is frequently browsed from the original data of "?", The attribute value of a variable (attribute name) "movie" of the agent characteristic is increased.

The agent characteristics include: 1) which field of information is preferred or disliked; 2) what information selection method is used; and 3) which information is not described in the property data of the agent from which field. 4) What kind of balance, cycle, and amount of information to provide, 5) How to change the agent characteristics itself, 6) How to express the information, such as how to edit the information. It has been described. These contents are described as a plurality of variables and subroutines.

If the relationship between the information on the network and the agent characteristics can be specified, the agent can collect, select, and edit information from a large number of information on the network according to the agent characteristics. If there is a function that describes the nature of information on the network with multiple variables and determines the relationship between those variables and the variables of agent characteristics,
The relevance between the agent characteristics and the information on the network is determined, collected and selected according to the selection method subroutine described in the agent characteristics, and can be edited by the editing subroutine. As a specific example of a method of automatically obtaining a variable that defines the character of information, if the information is described in HTML, it is selected based on the frequency of appearance of a keyword and the frequency of appearance in a heading, and the keyword, the frequency of appearance, and the like are information. It is possible to use a variable that defines the character of If described in XML, the nature of the information can be defined as a home page element (hereinafter, referred to as an HP element) and described using variables such as sub-element names, attribute names, and attribute values. Furthermore, a method is conceivable in which the relevance of variables such as keywords (sub-element names and attribute names) and values is determined by defining a function such as creating a matrix that is previously quantified and symbolized. If the relevance is high, it is estimated that there is a high probability that the information matches the needs of the viewer and the wants. By preparing a plurality of such matrices, it is also possible to create variations of the selection method. If we can identify the relationship between agent characteristics and information on the network,
After that, information can be selected and edited according to the selection and editing methods described in the agent characteristics.

[0011] When selecting information to be provided to the viewer, data mining and other analysis and estimation methods based on the information extracted by the agent from other viewers as well as the information extracted from the viewer by the agent up to then. By distributing information that is likely to match the preferences of the viewer using, the potential needs of the viewer can be more efficiently located. In the above example, based on the result of data mining, the relevant information is selected by modifying the relevance numerical value on the matrix of the two sub-element names and attribute names that are found to be highly relevant. Increase the probability of

By repeating such a process, the agent will be able to display accurate information on the viewer's terminal, and the viewer will be able to talk on a daily basis, become familiar with his preferences, and provide information that is nifty. Will feel more attached to the agents who carry the, and will be used more frequently. Therefore, this system is an effective means for activating e-commerce on a small display screen terminal such as a mobile phone.

Although the agent has a function of estimating the wants of the viewer, the estimating method is not limited to one, and depending on the information selection method set for the agent, even if the agents have the same purchase history and information selection history, After that, they grow up differently.

Also, unlike the above existing method, the agent is a set of programs or data, so that the agent itself can have various applications. For example, it is also possible to accumulate the information obtained by the agent and construct and update a database unique to the viewer's agent. Further, it is possible to improve the information selection method using the database. A function of appropriately exchanging information with agents of other viewers can also be assumed. Even if the information cannot be displayed on the mobile phone, the agent can automatically send the information file to the viewer's home computer by e-mail, so that the information can be viewed slowly after returning home.

If the agent has a character such as a rabbit or cat and displays it on the screen of the terminal, the agent is recognized as a virtual pet such as a "post pet" or "Tamagotchi", and can be viewed in a manner similar to daily conversation. Attributes, preferences, and opinions on the displayed information can be more naturally derived like a game. If you continue to do so, you can obtain a fairly deep information for estimating the potential needs of viewers, and build a highly available information providing system that provides sensible information and is familiar to everyone it can.

For example, if a character has a strong curiosity and has a wandering habit, random information irrelevant to the preference of the viewer is frequently carried. While interacting with these characters, viewers may come across unexpectedly interesting information. If the character is a science boy, it will collect a lot of relatively easy-to-understand scientific information. If the information provided does not match the viewer's preferences and the viewer acts on the agent, the agent may make a more secure choice for a while. For a while, they may try to provide random information. Viewers
If you send an e-mail to a friend's agent who also has an agent and get them to get along, they will exchange their preference information and will be affected and return. Agent characterization not only smooths out the viewer's attributes and preferences like a questionnaire in the form of dialogue with the character, but also, as described above, the viewer's wants with more advanced game characteristics. Is a powerful way to bring out Build a virtual everyday environment such as a clinic that changes the information selection method, a coffee shop that exchanges information with a friend's agent, and a marriage that obtains the information selection ability of a friend's agent. Become. As the reaction of the viewer, in addition to the information browsing status, the selection result of the indicated option, arbitrary input, purchase history of goods and services, and the like, the site is specified by symbol input as in Japanese Patent Application No. 11-105677. By combining the services, it is possible to easily estimate the preferences of the viewer with high accuracy.

[0017]

DESCRIPTION OF THE PREFERRED EMBODIMENTS In the following example, most of the functions of the agent are placed on the system operator (server) side in order to minimize the program of the agent mounted on the terminal (client). Therefore, the agents are distributed on both sides of the client and the server, and are identified by the agent name or the IP address and handled as one. However, here, in order to explain from the physical configuration, for convenience, the program and data on the terminal side are called agents,
The part provided on the side of the operator of the information providing system according to the present invention (hereinafter simply referred to as system operator) will be described as being included in the functions of each program and server. The terminal 11 is an Internet information terminal device for a viewer to view information. The agent 12 is an agent (program) that is placed on the terminal 11 and displays information received as an e-mail, and automatically creates and sends data input by a viewer and other data as an e-mail.
It is. The mail server 21 is a mail server that sends and receives electronic mail to and from an agent. The mail creation program 22 is a program that automatically performs a series of operations from receiving an email from the mail server 21 to creating an email to be sent to the agent 12. The agent DB server 23 is a database server on which agent characteristics of many viewers are set.
The WWW server 24 is a server connected to the Internet and storing a large amount of information like a mirror server. The subroutine group 25 is a server in which various subroutines called by the mail creation program 22 are placed. The matrix 26 is a program called by the subroutine group 25 and returns a matrix database and corresponding relevance values therefrom. The attached server 27 is a server in which a database and an application of each agent are placed. Mail server 21, mail creation program 22, agent DB server 23, WWW
Server 24, subroutine group 25, matrix 26,
The attached server 27 is placed under the control of the system operator 31.

Agent 12 on viewer terminal 11
Has the IP address information of the terminal 11, the image information of the character displayed on the screen, and the agent name. The agent 12 displays a character, displays an information file carried by e-mail, executes a conversation program, measures a viewing time of each information, orders a product by a viewer,
Creation and input of input screens such as selection of browsing information, evaluation comments, and browsing requests, creation of e-mails based on input or measurement data, transmission, (requesting, receiving, and displaying additional information to the server based on browsing wishes) ) And other functions.

The agent characteristics of each agent placed in the agent DB server 23 are described as a database composed of the following data. Browsing attributes such as agent name, IP address information, gender, age, etc., fields of interest, products, disliked information fields, application operation instruction information, etc., input from the viewer terminal, browsing time of each information Data collected by the agent 12 such as a history, a purchase history of a product, a list of already delivered information to a viewer, a desired information reception frequency, and a communication history with another viewer's agent are accumulated. Further, an agent characteristic determination subroutine is designated, and the above data is additionally written by the subroutine, and an agent characteristic variable calculated therefrom is given. These variables are made up of the degree of importance of each sub-element such as the information field sub-element of the HP element, the product name sub-element, the brand name attribute, and the product feature attribute, each attribute name, and the attribute value. The characteristic determination subroutine modifies the agent characteristic variable every time the data collected by the agent from the terminal is added. However, the characteristic determination subroutine can be temporarily fixed so as not to be corrected. The situation in which data such as these sub-elements, their importance, and attributes are being modified is a process of gradually finding out the needs and wants of viewers, so it can be expressed as "learning" and "growing up". I can say. For example,
If the information viewing time for a specific sub-element name or attribute name is long,
The degree of emphasis increases, and the degree of emphasis decreases if it is often discarded without being viewed. The specification of an information selection subroutine that determines an information selection method and its transition are described, and an editing subroutine that specifies an editing method is specified. Attached server 2 of the agent
The addresses of the database and application in 7 are also described.

The system operator 31 manages information on the network. The first management task is to standardize the description method of the information unit. When XML is assumed, the character of the information is specified, such as information field sub-elements of HP elements, product name sub-elements, brand name attributes, product characteristic attributes, and other sub-elements, their attribute names, and attribute value setting criteria. Standardize the description method. By this standardization, each information is characterized in a unified system. Furthermore, a method for measuring the relevance between the agent characteristic variable and the character of the information on the network is defined. As an example, a matrix that defines the depth of the relationship between the information field sub-element, the product name sub-element, the brand name attribute, the product feature attribute, and other sub-elements and the attribute names is created. With this and the attribute value, it becomes possible to numerically express the degree of relevance of the information for which the attribute is determined. A specific image of this matrix is that, for example, the value at which the attribute name “bag” and “shoes” of the product name sub-element intersect on the matrix has a value that is highly relevant to 100, and “bag” and “fishing” are described. For tools, it will be as low as 10. In addition, an association matrix between attribute names of different sub-elements may be created. For example,
A matrix is conceivable in which the female attribute of the viewer sub-element has a higher numerical value of the association between the attribute name "bag" and "shoes" of the product name sub-element than the male attribute. In addition, it is desirable to describe the creation date of the information in the nature of the information.

Second, when various information on the network has no attribute set, the attribute name of each sub-element of the HP element is obtained from words, contexts, images, etc. contained in the text in the information. A program for estimating and setting values is prepared, and attribute names and values are set in the information. The simplest method is to set an attribute name and an attribute value based on the frequency of occurrence of a word.

The system operator 31 prepares means for the agent to collect, select, and edit information. The first is to prepare multiple agent templates.
When the agent is initialized, a copy of this template is used as an agent characteristic in the agent DB server 23.
Is stored in

Second, several types of information selection subroutines specified by the agent characteristics are prepared. This information selection subroutine is a program that measures the depth of relevance between the attribute names and attribute values of each sub-element such as the information field sub-element and product name sub-element of the agent characteristics and the nature of each information on the network. It determines how strictly the depth of sexuality is to be applied for selection, what percentage to select without applying, and what to do with the combination ratio of each information according to the importance of sub-elements. According to this determination method, various information selection subroutines are created, for example, giving priority to the latest information, giving importance to sub-element names to obtain product-centered information, and obtaining relatively related information even if attribute names are different. . The information selection subroutine specifies the information to be passed to the agent based on the sub-element name passed from the agent characteristics, its importance, the attribute name, and the attribute value as external variables and based on the relevance shown in the matrix of section 0020. That is, it is a program for realizing the function of selecting information by the agent. The content of the subroutine is arbitrary, but the simplest is a program that determines a bunch of information composed of a certain amount of the same information as the sub-element names and attribute names of the agent characteristics in proportion to the importance of each sub-element. is there. A subroutine for adding information that has a good reputation to a viewer having the same attribute from among the latest information can be assumed.

Third, a new relation between the agent characteristics and the preferred information is discovered by a technique such as data mining, and the matrix of 0020 is modified based on the result, or some information selection subroutines are used. The purpose is to provide a more convenient information selection means for the viewer to reflect.

The fourth step is to prepare a characteristic determination subroutine for modifying a plurality of agent characteristics. Various information that agents have carried by e-mail is Agent D
At the same time as adding to the agent characteristic of the B server, the characteristic determination subroutine specified by the agent characteristic is called, and the agent characteristic variable is modified.

Fifth, an editing subroutine for creating a plurality of electronic mails is prepared. By selecting this subroutine, the layout on the terminal, the conversation language, and the like can be selected. In addition, an index and a conversation program with the character are added.

As described above, the characteristics of the information on the network are described, the relevance (compatibility) between the information and the characteristics of the agent is measured, the method of selecting the information is determined, the information to be collected is specified, and the e-mail is sent. Created.

The mail server 21 stores respective electronic mails created by the agent 12 and the mail creation program 22, and sends and receives them. When an e-mail is received from the agent, the e-mail is automatically transferred to the e-mail creation program 22 with the description of the agent name as a trigger.

The mail creation program 25 has a main program from receiving an email from the mail server 21 to creating an email to be returned to the agent 12 and sending it to the mail server 21.

A large number of Web sites are collected on the WWW server 24, and each home page (hereinafter simply referred to as HP) is described in XML. Each HP has one or more sub-elements, attributes, and attribute values such as an HP element indicating the characteristics of the HP, an information field sub-element, a product type sub-element, a brand name attribute, and a price attribute. (Hereinafter, simply referred to as sub-elements and the like.) The WWW server 24 collects information files by appropriately traveling to other WWW servers on the Internet in addition to the information files transmitted externally, and collects a large amount of information. Has accumulated. If the sub-elements and the like standardized in the HP are not set, the sub-elements and the like of the HP element are searched and estimated from the tags and keywords in the XML document.
An HP element is added and described in the XML document of the HP.

The subroutine group 25 includes one or more characteristic subroutines, information selection subroutines, and editing subroutines.

In the matrix 26, a plurality of matrices describing the depth of association between the sub-elements of information on the WWW server and the sub-elements of the agent characteristics are placed. In response to a call from the subroutine group 25, Returns the relevant relevance figure.

The attached server 27 stores a database and an application of the agent. This is presumed to be a function of transferring an e-mail to a personal computer at home and of building a database unique to an agent (viewer).

The viewer, after initially setting the model of his agent on the terminal, sends the e-mail to the server as it is.

A mail server 21 that receives an e-mail describing only the agent name and IP address information
Transfers an e-mail to the mail creation program 22, triggered by the description of the agent name. The e-mail creation program that received the e-mail
It asks the agent DB server 23 that the agent characteristics are not set from the agent name and the IP address, and copies the corresponding template of the agent characteristics in the server to the agent DB server 23 as the agent characteristics of the agent. save. Further, a predetermined editing subroutine is called from the subroutine group 25, a standard initial stage questionnaire file is created as an electronic mail, and sent to the mail server 21. The agent 12 delivers this to the viewer and displays it.

The agent 12 displays the character on the screen of the terminal and inquires about the questionnaire items of the e-mail, for example, attributes such as sex and age, hobbies, interests, and desired functions. The agent composes an e-mail based on the answer of the viewer and sends the e-mail to the mail server 21.
Deliver to

The mail server 21 checks the description of the agent name in the electronic mail, and checks the mail creation program 2
Transfer to 2.

The mail creation program 22 calls the agent characteristics of the agent DB server 23 based on the result of the questionnaire, adds the data, calls the specified characteristic determination subroutine from the subroutine group 25, and calls the sub-elements and the degree of importance. For example, an agent characteristic variable is calculated, added to the agent characteristic, and stored in the agent DB server 23. The above operation may be performed a plurality of times.

The mail creation program 22 calls the designated information selection subroutine from the subroutine group 25 based on the calculated agent characteristic variables, and extracts appropriate information from the information stored in the WWW server 24. At this time, the called information selection subroutine
The matrix 26 is appropriately called. In addition, information corresponding to a sub-element or the like designated as dislike in the agent characteristics is automatically excluded. Further, the mail creating program calls the designated editing subroutine from the subroutine group 25, creates an electronic mail by adding an index, a questionnaire, a greeting, etc. to the extracted information as appropriate, and sends it to the mail server 21. When creating the electronic mail, the mail creating program 22 may change the index and the link setting of each information in order to enhance the convenience.

The agent 12 receives the electronic mail from the mail server 21 in accordance with the designated receiving frequency, and displays it on the terminal 11. The agent 12 automatically measures information such as whether or not each information file in the received e-mail has been opened, the browsing time, and the like, and further evaluates the viewer, responds to a questionnaire, and purchases a product or service. Create an email that summarizes the purchase target, quantity, payment method, etc. The e-mail is sent to the mail server 21 automatically by the operation of the viewer or automatically sent to the mail creation program 22 by being received from the agent 12. When the agent 12 sends a mail to another agent, the mail server 21 also notifies the mail creation program 22 of the situation.

The mail creation program 22 extracts data to be stored in the agent characteristics from the reply result of the e-mail, calls the agent characteristics data from the agent DB server 23, adds the extracted data to the agent characteristics, and further writes a subroutine. Call a characteristic determination subroutine from group 25, agent characteristic variables,
That is, the degree of importance, sub-elements, and the like are updated and corrected.

In this process, a learning ability is created, in which, for example, the potential desire target is narrowed down or the situation changes gradually from the repetition of the viewer's reaction.
The other is the ability to learn from readers' reaction to information sent randomly or selected by methods derived from analysis such as data mining, and discover that they are interested in unexpected fields. Is added. Further, it is possible to enhance the learning ability by transplanting characteristics of another agent's agent, particularly a friend's agent, that is, by adding learning results. this is,
It also has the effect of providing a common topic with friends.

Hereinafter, items 0039 to 0041 are repeated. In this process, the mail creation server 22
It is a program common to all agents and plays a role like traffic control. The individual characteristics of each agent are all condensed in the agent characteristics stored in the agent DB server 23. The information to be selected is specified by specifying an agent selection variable and an information selection subroutine in the agent characteristics.

In the above-described process, the attached server 27 appropriately executes the mail creation program 2 according to the description of the e-mail.
2, the corresponding database of the attached server 27 is updated, or the designated application operates.

Unless maintenance or data mining is performed, the system operator 31
Server 23, attached server 27, mail creation program 2
If the system cannot be accessed to extract the characteristic data of the individual agent, the privacy of the individual viewer can be protected from the system operator. In performing the maintenance and the data mining, a method in which personal data of individual viewers cannot be extracted is used.

When it is desired to copy the characteristics of an agent to the characteristics of another agent, a characteristic determination subroutine for duplicating the characteristics is called with the understanding and instruction being described in the e-mail as a trigger, and the agent characteristics of the copy destination are called. , It is sufficient that the agent characteristic of the copy destination is overwritten on the agent characteristic. In the case of affecting a friend's agent, call the characteristic determining subroutine that also affects the friend's agent, confirm that there is a statement of consent to be affected by the agent characteristic of the other party, The agent characteristics may be modified based on the calculation formula.

The information selection subroutine for estimating the wants and preferences of the viewer, the characteristic determination subroutine for modifying the characteristics of the agent, and the editing subroutine may have various variations, and some of them may be combined. For example, if you buy a bag, introduce other products such as belts of the same brand, select information according to the type of the viewer by some data mining method from the information of many viewers, in the same area Various methods may be employed, such as providing information preferred by other viewers of the same attribute as the viewer. Therefore, it may be arbitrarily created according to the purpose. It is also conceivable that the characteristic determination subroutine automatically changes the designation of the information selection subroutine, the editing subroutine, or the characteristic determination subroutine itself by accumulating constant data of agent characteristics. The description language is arbitrary, such as Java or C language. In the above example, a matrix is used as a method for determining the relationship between the information on the network and the agent characteristics, but any method may be used as long as the relationship can be specified. Also, as long as the degree of relevance of the information can be determined, the matrix may be created in any manner.

In the construction of the present system, in the above-described embodiment, the information is electronic mail, the Internet, or the like, but other communication means may be used. Further, a network via the Internet and another mail server between the agent 12 and the mail server 21 may be used.
Further, the broadcast between the mail server 21 and the agent 12 may be used as a part of the downstream distribution means. HTM
Languages such as L and XML do not matter. The terminal to be a client may be any device such as a personal computer, a mobile phone, or a TV, and the server may be located at a provider's server or a personal Internet connection server, or may be connected via a separate line from there. It doesn't matter.
The network used is the Internet, an intranet, or any other network. In the above example, the input means of the viewer is assumed to be text input, button operation of a remote controller, or the like, but voice and other input means may be used in combination.
Although the agent name and the IP address are used to specify the agent, the agent may be specified by any other code. The location of the agent may be resident at the terminal of the viewer, or may be called from the server each time it is used. Further, the agent may be a mobile type, a fixed position type, or may be distributed and located on the client and server sides. Whether or not a character is set is optional, and the agent on the client may be only a set of data. The program description language of the agent is also arbitrary such as Java or C language.

[0049]

As the agent acquires a lot of information through long-term use of the viewer, the agent gradually grows, and it becomes more successful to find the viewer's wants more accurately. This can be described as growing up or being customized. If the agent becomes proficient in finding the wants of the viewer, it will provide a more effective sales promotion method in e-commerce on the limited display screen of the viewer who has only vague needs. Furthermore, the result of data analysis based on the attributes, preference information, purchase history, and the like of viewers collected naturally will be important marketing materials for the company.

By copying this agent to the terminal of another viewer, the agent raised by the original viewer can be extended to other viewers. The original agent, who became a celebrity pet, could itself have economic value.

The first feature of the present invention is that the agent has a learning ability. In the process of daily browsing of information by the agent, the agent uses the input of the viewer, information browsing status such as information browsing time, product and service purchase history and purchase target attributes, selection method selection, transition, or agent The information selection method is modified by automatically changing the selection method triggered by certain data accumulated by
The estimation ability can be improved. Also, if information based on the characteristics of the viewer is provided based on the results of analysis such as data mining, it is highly likely that even information in fields that the viewer has not entered is sensible information, If there is a positive response, the agent characteristics are modified so as to emphasize information in the field. It can be said that such a process is capable of learning. In such a process, the ways in which learning abilities are enhanced are very diverse. Furthermore, while reflecting the will of the viewer, the learning ability is enhanced even in a place where the viewer is not conscious. It can also be said that individuality is generated in the agent by learning.

The second feature is that the agent operates independently, not by the operation of the viewer. The agent performs the work of automatically collecting, selecting and editing information by estimating the needs and wants of the viewer, regardless of the operation of the viewer. In addition, in addition to the viewer's input, various data such as the measured value of the browsing time of each information, the attribute data of the purchased product, and, in some cases, the input from the sensor attached to the terminal are automatically accumulated to enhance the estimation ability. To go. In this regard, the agent can be said to be independent of the viewer. In addition, the information selection means such as the matrix described in the section 0020 is appropriately corrected based on the result of data mining or the like, so that it can be said that the viewer is independent from the viewpoint. In addition, the viewer sometimes inputs some answer to the information displayed by the agent, but as a result, the viewer is not always conscious of how the agent will change, which may cause unexpectedness.

The third feature is that the agent is independent of the system operator. As described above, the agent modifies the information selection method based on various data obtained from the viewer. This data includes an input intended by the viewer, such as selection of an agent model, and the information selection method is modified where the system operator cannot control. This is different from the conventional information selection method set by the system operator to customize the provided information. The system operator only provides the framework of the system according to the invention. In addition, the system operator may use an agent DB server, attached server,
If the system does not have access to some subroutines and cannot extract the characteristics of individual agents, the state of the agents and the learning process cannot be known by the system provider, and the method of selecting information for individual agents cannot be controlled. In this respect, it can also be said that it is independent.

The fourth feature is that an open information providing system that does not specify an information provider can be constructed. If the system operator cannot access the agent DB server and the attached server, and the ownership of the data (copyright) on those servers is the viewer, privacy can be protected. Operators are more likely to get consent to provide comprehensive information from reassured viewers. If you have comprehensive consent to provide information, there are many companies that want to provide information,
An open information provision system not limited to a specific company can be constructed. If the system operator concludes an information search proxy agreement with the reader, this information provision system is only a position to provide a framework for the reader to select and collect information, and information search sites such as Yahoo There is a possibility that it will be used instead of. If the data belongs to the viewer, the agent can transfer (copy) the data to another person at his / her will. The third and fourth points are points that cannot be realized even if a conventional system, for example, an information providing system for customizing information on the server side is developed.

Fifth, a viewer as described above,
Agents that are independent and unique to both system operators are dedicated to individual viewers and have characters.
The personality setting of the character sets the degree of the gap between the viewer's assumption and the display frequency in the field of the displayed information, the manner of reacting to the user's answer, and the like. You can also set the display language such as honorifics and tamed words, the display image of the character, and the way to edit the collected information such as index settings. As a result, the amusement can be significantly improved as compared to simply providing and inputting information mechanically, and the effect of increasing the frequency of using the information providing system will be brought.
This setting is optional for the viewer, and a program that influences the character setting of the character of a friend who frequently exchanges e-mails is also possible.

As a result of the above-mentioned features, the information providing system according to the present invention uses an information terminal device having a small display screen such as a portable terminal, or uses an interface such as a remote controller which makes it difficult to input arbitrary characters. In this case, when trying to obtain desired information with a small number of keywords (vocabulary) or ambiguous intentions, an effect of quickly providing information that appropriately captures the needs and wants of the viewer is provided.

The first effect is that, through daily use, input to the provided information, browsing time, shopping history, and other browsing information are automatically accumulated, and the conscious or unconscious information of the browsing is obtained. The point is that data relating to selection preferences is accumulated, and information that each viewer likes can be provided more accurately. In the past, it was troublesome to answer many questionnaire items at first, and the way of providing information was uniform, and it was difficult to construct an information providing system that reflects the preferences of each viewer. If the viewer selects and sets a relatively preferable model from among several types of agent models at the beginning, the agent gradually grows into an original agent that meets the personal preference of the viewer. In addition, through daily conversations with the characterized agents, viewers will be able to grow their agents without any hassle,
The growth process itself can be fun for the viewer.

The second effect is that if the information providing system is modularized as in the embodiment, the system operator gradually evolves the information providing system and gradually increases the learning ability and the information selecting ability of the agent. It is a point that can be raised. In the embodiment, the agent uses a plurality of prepared subroutines for determining, selecting and editing information, using the agent characteristics which the agent has individually as variables, so that the system operator can provide additional subroutines which are more excellent. For example, the ability of the agent to learn, select information, and express is enhanced. It will also be improved by improving the method of characterizing information on the Internet and improving the matrix. It is conceivable that the viewer saves a unique subroutine in the attached DB server.

A third effect is that the agent has a possibility to grow in a direction unexpected to the viewer, but in a preferable direction. Agents often carry randomly selected information to a certain extent, depending on their personality settings. It also carries information selected based on the results obtained by data mining of a large number of viewer data and the characteristics of the viewers. These may or may not be desirable to the viewer, but they give the viewer an opportunity to discover their preferences that were completely unconscious. If you build a program that modifies the method of selecting and editing information triggered by the accumulation of a certain amount of data, the viewer will still be somewhat surprised.

The fourth effect is that the viewer can own his own (agent) database and application. Specific examples of this effect are shown in section 0014.

A fifth effect is that the viewer voluntarily inputs his / her own preference data continuously for the purpose of raising an agent. If the action inputted by the viewer is performed in accordance with the purpose of the company, it becomes a complicated operation, and if the same action is performed in accordance with the purpose of the viewer himself, it is fun. As a result, a company that provides information can be expected to accurately transmit information to viewers who have strong needs and wants, even if it is not possible to know in advance which viewers the information will be transmitted to.

The sixth effect is that the privacy of the individual can be protected from the system operator. As a result, not only is it convenient and enjoyable, but also the viewer can use this information providing system with peace of mind, and as a result, the system operator has the expectation that a large number of viewers can be obtained. In addition, it is more likely that a reader will obtain a comprehensive information provision agreement that does not limit the information source, and the system operator will be more likely to conclude information provision contracts with many companies. If the viewer and the system operator enter into a proxy agreement for information selection, the system operator is merely performing the information collection agency business of the viewer, and the viewer can use the information on the entire Internet, in other words The system operator can provide all information on the Internet as well as a robot-type search site. The company that provides the information cannot obtain information about the individual who browses this system information, but personal information can be obtained if the visitor directly accesses the company using the information obtained from the information providing system. It is similar to the existing Internet.

The seventh effect is that the agent can transfer. Viewers can take over agents raised by other viewers and easily use competent agents. As a result, the agent itself can be a product, and communication and amusement such as synthesis and exchange with the characteristics of a friend's agent also occur.

The information is not limited to text and still picture information such as homepages, but also moving picture information such as television programs and audio information. If broadcasting is used as one of the down lines of the network connecting the terminal of the viewer and the server, the agent may have a function of distributing a large amount of information. In such a case, information such as a time zone and a channel in which the corresponding program is broadcast by e-mail or the like is distributed, and the agent automatically extracts the information and sets a download reservation in a home server or the like. I just need. In addition, the agent may have a function of extracting the response of the viewer (viewer), such as whether or not the user has viewed the entire recorded program, and automatically writing the response in an e-mail to be created. Furthermore, if the characteristics of programs such as television and radio broadcasts are standardized and determined, and the agent characteristic variables are determined from the reaction data of the previous viewer, a very sensible automatic recording function can be realized.

As a result of this effect, the present information providing system successfully picks up a product that is likely to be liked during shopping from a large number of product information, and attracts the viewer's interest during education. In addition, we will introduce information according to the level of the viewer sequentially, and if it is a business, share the method of extracting information of excellent employees among all employees, etc.
In the world of the Internet, which is likely to be swallowed by the flood of information, it is a means of providing useful information even when using a terminal with a small display screen, operating with a remote control, or even those who are not good at computers.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of an apparatus.

FIG. 2 is a diagram showing a flow in which the system operates.

FIG. 3 is a diagram showing an example of a matrix representing the relationship between the personality of information and agent characteristics.

Claims (8)

[Claims] 1. A conscious and subconscious information selection criterion and selection method of a viewer who wants to select and collect a certain amount of information from a large number of information existing on a network and estimate the information. A terminal device that displays, to a viewer, information collected, selected, and edited from a large number of specific or unspecified information existing on the network based on the selected information selection criteria and selection method (hereinafter, simply referred to as estimated information selection criteria), A device that combines a server and software to determine criteria for selecting estimated information based on data such as viewer input, usage of display information, and selection of the information selection method itself. By repeatedly acquiring and accumulating data such as the usage status of information, selecting information selection methods, etc., and estimating information selection criteria and selection methods of viewers gradually by modifying the estimated information selection criteria Allowing will increase the force and the device. 2. The apparatus according to claim 1, wherein the image of the character is displayed on the terminal of the viewer, the information is displayed on the terminal device with the setting that the character has been carried, and the viewer is set with the setting of dialogue with the character. A device that prompts the user for input. 3. An information providing system for providing a viewer with information collected, selected, and edited from information on a network, using the device according to claim 1. 4. An information providing system using the device according to claim 2, wherein information collected, selected, and edited from information on a network is provided to a viewer. 5. The system according to claim 3, wherein
Except for the operation, maintenance, and data mining operations of the described system, the operator of the information providing system must not store any personal data of the viewer, such as data entered by the viewer, data specifying how to select information, or An information providing system that restricts access to applications containing personal data and protects the privacy of each viewer from those who operate the information providing system. 6. The system according to claim 4, wherein
Except for the operation, maintenance, and data mining operations of the described system, the operator of the information providing system must not store any personal data of the viewer, such as data entered by the viewer, data specifying how to select information, or An information providing system that restricts access to applications containing personal data and protects the privacy of each viewer from those who operate the information providing system. 7. A method of selling goods and services that promotes sales by displaying the goods and services by estimating consumer needs and wants using the system according to claim 6. 8. An educational system using the system according to claim 6, wherein the learning status of the viewer is grasped, and information is transmitted according to the status of each viewer, thereby improving the learning effect.