KR20060113312A - Method and system of proposing products and contents through analysing patterns of conversation data - Google Patents

Abstract

A method and a system for recommending preferred goods/contents through conversation pattern analysis are provided to recommend the preferred goods/contents by storing conversation between the conversational agent and the user, and analyzing the conversation. A conversational agent service server(4) provides a service for offering the conversation between the user and the conversational agent. A conversation database(5) is linked with the conversational agent service server and stores question/answer pairs. A conversation analysis/recommendation server(7) analyzes preference of the user by analyzing the conversation contents between the user and the conversational agent, and recommends the goods or the contents based on the analyzed preference.

Description

METHODS AND SYSTEM OF PROPOSING PRODUCTS AND CONTENTS THROUGH ANALYSING PATTERNS OF CONVERSATION DATA}

1 to 4 are diagrams illustrating a conventional conversation agent service,

5 is a view for explaining an example of a system according to the present invention;

6 is a diagram illustrating an example of contents of a conversation between a user and a conversation agent;

7 is a view showing an example in which a user's preferences are analyzed and organized according to the present invention;

8 is a view showing an example of recommending a product or content based on the user's analyzed preferences according to the present invention;

9 is a diagram illustrating an example of a process of recommending a conversation agent to a user according to the present invention.

The present invention relates to a method and system for recommending a preferred product and content by analyzing a pattern of a conversation, and in particular, analyzing a content of a conversation between a user and a conversation agent, analyzing a user's preferences, and based on the analyzed preferences. The present invention relates to a method and system for recommending a preferred product and a content through a pattern analysis of a dialogue recommending the content.

A conversational agent is a chat robot that means a software agent that can talk to people. It is called chatterbot, chatbot, chatterbox, etc. (see Wikipedia Encyclopedia [Webpage] Address: http://en.wikipedia.org/wiki/Chatterbot]).

1 is a view for explaining an example of the prior art, A.L.I.C.E. Alice is a conversation agent provided by the Artificial Intelligence Foundation through the website www.alicebot.org. The chat agent Alice provides a preset response to the question "what is a chatter botterbot".

FIG. 2 is a view for explaining another example of the related art, and shows a manner in which a dialogue agent Lingubot, which kiwilogic.com provides through a website www.elbot.com, leads a conversation by first presenting a set theme.

3 is a view for explaining another example of the related art, which shows a screen for chatting with a chat agent (email address: simsimi200@hotmial.com) on an msn messenger, and this service is manufactured by ismaker website www.simsimi.com Provided through.

In addition, various chat agent services are available on the web, such as chat with various chat agents through www.personalityforge.com and with chat agent ella through www.ellaz.com.

Such a chat agent service is a chat between a user and a chat agent, that is, a chat robot, rather than a chat between the user and the user, and mainly by accumulating the expected question / answer pairs in advance. Through pattern matching), the user's questions can be answered. For example, if a user asks "Have you eaten?", The user goes through the appropriate natural language processing and chats with the user by finding the corresponding response "No yet" from the conversation DB where the question / response pairs are stored. Will be.

4 is a view for explaining another example of the related art, where the conversation agent simsimi of FIG. 3 recommends a product (eg, a network-related product) to a conversation counterpart. However, since the recommendation is unilaterally suggested by the conversation agent service provider or the product advertiser regardless of the preference of the conversation partner, it is insufficient to attract the conversation partner's attention or increase the effect of the product advertisement.

On the other hand, Japanese Laid-Open Patent Publication No. 2003-248681 is a method of analyzing a user's preferences from a given text using a natural language processing technique. A method of finding out the user's preferences by comparison is presented.

In view of the foregoing, an object of the present invention is to provide a method and system for recommending a preferred product and content by storing and analyzing a conversation between a user and a conversation agent.

To this end, the present invention provides a first step of transmitting information of a user who wants to talk to a chat agent to a chat analysis / recommendation server. And a third step of receiving the information transmitted by searching for the content, and a third step of expressing the information transmitted by the chat analysis / recommendation server through the chat agent. Provide recommendations.

In addition, the present invention provides a conversation agent service server that provides a service that can communicate with a user through a conversation agent, a conversation DB interoperating with a conversation agent service server and storing question / answer pairs, and a conversation content between the user and the conversation agent. It provides a preference product and content recommendation system through a pattern analysis of the conversation, characterized in that it comprises a dialogue analysis / recommendation server that analyzes the user's preferences, and recommends goods or content based on the analyzed preferences.

Hereinafter, the present invention will be described in more detail with reference to the drawings.

5 is a diagram illustrating an example of a system according to the present invention, in which the present invention is a user computer (1, 2), the Internet (3) as a network, a conversation agent service server (4), a conversation with stored question / answer pairs In addition to the DB 5, it further includes a conversation storage DB 6 storing conversation contents between the user and the conversation agent and a conversation analysis / recommendation server 7 analyzing the conversation stored in the conversation storage DB 6; . The user can connect to the chat agent service server 4 via the Internet 3 via the user computer 1 and have a conversation with the chat agent provided by the chat agent service server 4. At this time, the response to the user's question is sent by the dialog agent service server 4 to the user computer 1 by finding the appropriate response from the DB 5 in which the question / response pairs are stored. Meanwhile, the conversation contents between the user and the conversation agent are stored in the conversation storage DB 6 and analyzed through the conversation analysis / recommendation server 7. The goods or contents related to the analyzed preferences or preferences may be presented to the user or communicated to the user through e-mail in a subsequent conversation with the conversation agent.

Hereinafter, a method of analyzing the user's preferences or preferences from the contents of the conversation between the user and the conversation agent will be described.

An example of analyzing a user's preferences or preferences according to the present invention is described from this dialogue. Various methods may be used by those skilled in the art in analyzing preferences (e.g., Japanese Patent Laid-Open No. 2003-248681, cited in the prior art), and should be understood as one example described herein. When talking about your preferences or preferences in conversation, you usually use the words "good" and "like" such as "I like apples" and "I like baseball." Therefore, the present invention proposes a method of analyzing a user's preference from the contents of conversation using "likes".

FIG. 6 is a diagram illustrating an example of a conversation between a user and a chat agent, in which the user (dialogue [visitor]) says "hello, the weather is cloudy", "what is a hobby", "I like baseball "Three sentences," the first sentence, which does not contain "likes" after the natural language processing, are removed, and the third sentence shows that the visitor likes baseball. This analysis of user preferences or preferences can also be done by the conversation agent presenting a predetermined question to the user and organizing the answers from the user.

 FIG. 7 is a diagram illustrating an example in which a user's preference is analyzed and organized according to the present invention, and it is indicated that the user has a preference for baseball of the user (dialog [visitor]), and preferably various Symbols can be managed collectively for a specific user (for example, the user's preferences for baseball, stakes, and spaghetti are collectively managed), and can also be counted when the user's conversations are continuously analyzed and mentioned several times. Symbols (e.g. user mentioned 1 baseball, 1 stake, 3 spaghetti) and categorized and managed to categorize certain symbols (e.g. put baseball in sports category, steak in food category) Management can be done efficiently. In addition, by managing additional information, for example, an email address, to the user as needed, it is possible to obtain convenience in recommending products or contents related to preferences to the user.

FIG. 8 is a diagram illustrating an example of recommending a product or content based on a user's analyzed symbol according to the present invention. On the left side of FIG. 8, a symbol analyzed by a conversation agent when the user (dialog [visitor]) visits again. Based on this example, a response (eg, "Visitor today is XX baseball game") is presented, and the right side of FIG. 8 is recommended to the user via messenger (eg, "Visitor today is XX baseball game"). ") Is shown.

9 is a diagram illustrating an example of a process of recommending a conversation agent to a user according to the present invention. First, the conversation agent service server 4 receives a user's conversation input (S901) and receives a conversation DB 5. While searching for a response (S902), the user information (e.g., a screen name) is passed to the conversation analysis / recommendation server (7) (S903), and the conversation analysis / recommendation server (7) received the information is stored in the conversation DB. (6) Or, if the user is analyzed by searching in a separate DB or not (S904), if there is no information about the user or no preference analyzed from the information as summarized in FIG. Is returned to the conversation agent service server 4 (S905), and if there is an analyzed preference of the user in the organized information as shown in FIG. Contents are separate D If there is no recommendation, it returns to the conversation agent service server 4 (S907), and if there is no recommendation, as in step S905. Information is returned to the dialog agent service server 4 (S905). When the information that there is no recommendation is returned, the conversation agent service server 4 expresses a response to the user's input retrieved through the conversation DB 5 as a response of the conversation agent in step S902 (S908), and recommends. When the content is returned, the conversation agent service server 4 may express the response of the conversation agent in various ways according to a predetermined rule, such as expressing only recommendation contents or expressions with the response retrieved through the conversation DB 5. (S908).

In the present exemplary embodiment, the above steps are performed by the user's conversation input in step S901. However, the process may be performed by requesting a conversation with the user's conversation agent instead of the user's conversation input. In this case, step S902. ) May be omitted.

According to the present invention, it is possible to provide a customized conversation agent service that can analyze contents of a conversation with a conversation agent and propose a product or content that is customized to a user's preference in a subsequent conversation.

Claims (2)

A first step of transmitting information of a user who wants to talk with a chat agent to a chat analysis / recommendation server; A second step of the conversation analysis / recommendation server finding a symbol analyzed from the user information, finding recommendation contents related to the analyzed symbol, and receiving the transmitted information; And, And a third step of expressing information transmitted by the chat analysis / recommendation server through a chat agent. A chat agent service server providing a service for talking with a user through a chat agent; A conversation DB interworking with a conversation agent service server and storing question / answer pairs; Preference analysis through pattern analysis of the dialogue, characterized in that it comprises a dialogue analysis / recommendation server for analyzing the contents of the conversation between the user and the dialog agent, and recommends goods or content based on the analyzed symbols Product and content recommendation system.

Images