WO2002101483A2

WO2002101483A2 - Network communication system

Info

Publication number: WO2002101483A2
Application number: PCT/SG2001/000116
Authority: WO
Inventors: Lung Hsiang Wong; Chee Kit Looi
Original assignee: Kent Ridge Digital Labs
Priority date: 2001-06-12
Filing date: 2001-06-12
Publication date: 2002-12-19
Also published as: WO2002101483A3; AU2001264526A1

Abstract

A network communication system has a number of user stations (2, 3, 4) which are connected via a data network to a remote chat server hosting chat room channels (5, 6, 7). Software is provided with a multi-agents structure that is able to sample the chat content of all channels. The software also makes recommendations to the individual users of the system concerning links to other users, chat channels and resources on the network by parsing the chat content and by making use of a keyword-based model of interest. The system is provided with a global chatter web server (1) that includes a generator for generating dynamic links between the individual user's stored profile. Links are generated on the basis of the real-time analysis of the user's chat, between the user's ongoing chat session and the other chat room channels in which related topics are treated in ongoing chat conversations or were discussed in previous chat conversations. The generation of dynamic links is independently of the specific system where the chat room channels are hosted, so that chat room channels from different chat servers in different parts of the world can be linked without a need for the user to log onto a new chat room channel to share part of their conversation.

Description

NETWORK COMMUNICATION SYSTEM

The invention relates to a network communication system wherein a number of user stations are connected via a data network to a remote chat server hosting chat room channels where text messages may be exchanged, comprising software with a multi-agent structure that is able to sample the chat content of all channels and in view of matching interests makes recommendations to the individual users of the system concerning links to other relevant users, chat channels or resources on the network by parsing the chat content and by making use of a keyword-based model of interest.

The Internet is a worldwide network communication system which enables users throughout the world to access information stored in the World Wide Web, to send and receive e-mails, to interact, to discuss issues, get assistance, learn and socialise.

Electronic mail or e-mail allows users to exchange text messages. The receiver of the electronic mail is able to receive the electronic mail at a time convenient for the receiver.

Therefore, the receiver is provided with convenience because the work being conducted by the receiver is not interrupted by the received electronic mail, as it would be the case by a call on the telephone. The sender is able to easily use the electronic mail without considering the time at which the mail is received, because the mail does not interrupt the receiver.

A chat system is, on the other hand, known as a means for communicating information in real-time. The chat system has a chat server which is a host computer and a plurality of terminal units connected to the same through the public telephone line or through data lines. A message received by the chat server from a terminal unit is immediately transmitted to other terminal units. Therefore, a chat system is suitable to allow users to discuss matters in a way which is similar to a phone call or a conference phone call. Real-time exchanges between people therefore are possible on the Internet through synchronous chat discussions. The difficulty in using chat rooms or chat room channels is to find out where groups of people are discussing topics in which the user is interested. Therefore, there is a need for tools to link people together according to their interests since online chat rooms tend to be isolated, which is not the case with web pages in the World Wide Web where linkages between pages are made possible by hyperlinks.

In order to overcome this difficulty, Neil W. Van Dyke, Henry Lieberman, Pattie Maes, Media Laboratory, Massachusetts Institute of Technology, have proposed in their article "Butterfly: A Conversation-Finding Agent for Internet Relay Chat" a network communication system in which the user's interface to IRC (Internet Relay Chat) is augmented with a software agent that deals with the information overload due to the many thousand potential conversations in which the user might be interested. The conversation-finding agent of the known system samples the content of the channels which are active in chat rooms of a given chat server and makes recommendations using a keyword-based model of interest. In order to deal with privacy, the known conversation-finding agent connects to an IRC server and appears as a normal user client joined to any channel it is sampling. While it is visiting a channel, the agent builds a term vector of occurrence counts of keywords it extracts from channel conversations.

A communication system which links chat room channels with web pages has been described in WO 99/44152. This publication describes a computer communication system wherein a number of user stations are connected via a data network to a remote server hosting a chat-space where text messages may be exchanged. A parsing system is provided which identifies keywords in the exchanged text messages and, using these as search terms sent to a further server hosting a search engine, obtains and makes available to the user auxiliary sources of data which may be of interest.

The United States patent 5,987,503 describes a system and a method for displaying an electronic mail messages related to a statement submitted in an online chat session. Statements submitted by participants in the online chat session are searched for a specific keyword. When a keyword is detected in a statement submitted by a participant in the chat session, mailboxes of the participants can be searched for electronic mail messages related to the specific keywords.

The problem underlying the invention is to provide for a network communication system which allows the user to have easy access to the wealth of information that can be found in chat rooms and web pages accessible via a data network, in particular the Internet.

This problem is solved in a network communication system according to the preamble of claim 1 by providing the system with a global chatter web server that includes a generator for generating dynamic links between the individual user's stored profile or, on the basis of the real-time analysis of the user's chat, between the user's ongoing chat session and the other chat room channels in which related topics are treated in ongoing chat conversations or were discussed in previous chat conversations, independently of the specific system where the chat room channels are hosted, so that chat room channels from different chat servers in different parts of the world can be linked without a need for the user to log onto a new chat room channel to share part of their conversation.

Since such a communication system is no longer limited to chat rooms of a particular owner or server, the user can interact with other similar minded much easier. The network communication system according to ^'the invention not only allows real-time exchanges between people through online chat, but also provides a tool to link people together with like-minded interests. The user receives recommendations on other chat rooms with ongoing conversations on similar or related topics. Besides that, the user can receive recommendations on web sites and other resources on similar or related topics. One advantage of the invention is that it allows the user to specify his profile and update it. The user also receives recommendations on chat rooms with previous chat conversations on similar or related topics, and the day and time when such conversations occur. Further details and advantages of the invention will become apparent from the following detailed description and the claims. The invention will now be described by a way of example with reference to the accompanying drawings in which:

Fig. 1 illustrates the architecture of a global chatter web server;

Fig.2 shows an internal schematic for the chatter web server;

Fig. 3 illustrates an embodiment of a user interface for a personal chat agent;

Fig. 4 shows a conversation in a virtual-virtual chat room;

Fig. 5 illustrates connections between personal chat agents, chat room agents, chat rooms and the chatter web server;

Fig. 6 illustrates the process of user-to-user /-room /-web site matchmaking; Fig. 7 illustrates the process of creating a virtual-virtual chat room;

Fig. 8 is a block schematic diagram for the process of launching an eavesdropping service;

Fig. 9 is a flow chart for illustrating the process of launching an eavesdropping service;

Fig. 10 is a flow chart of user-to-user /-web page /-room matchmaking performed by the chatter web server;

Fig. 11 shows a flow chart of user profiling performed by the personal chat agent; and

Fig. 12 illustrates the room-to-room matchmaking and the virtual-virtual chat room creation which is performed by the chatter web server.

As it can be seen in Fig. 1, the network communication system consists of three main components which together create a chatter web with an agent-based architecture in view of a global web of chat services with intelligent features.

The first of the three components is a global server 1 , which is hereafter called chatter web server or CWS. The agent-based architecture of the network communication system according to the invention which can be called chatter web also comprises as second component a number of personal chat agents 2, 3, 4. The personal chat agents or PCAs are agents sitting on a user's desk top as a stand alone application or as a web-based client. The personal chat agents serve as an intelligent chat client when the user logs on to any chat room. Fig. 1 shows three external chat rooms 5, 6 and 7 which are called chat room 1 , chat room 2 and chat room 3. The third component of the architecture of the chatter web according to the invention are chat room agents. Fig. 1 shows a first chat room agent 8 and a second chat room agent 9.

The agent-based architecture of the network communication system shown in Fig. 1 is based on a community of agents approach in which every user can employ the services of his personal chat agent. Besides that, a user with the name Inky can use his personal chat agent 2 to leverage on an other user's agent to perform a multitude of tasks. A user can specify what services he desires in his chat conversation such as filter spamming, produce logs of conversations, measure individual participation and moderate language use.

In order to protect the privacy of the users, the users Inky, Dante and Tintin shown in Fig. 1 can also specify their current chat room conversation to be private, i. e. not to be eavesdropped by other agents.

Fig. 1 illustrates also a possible interconnection between the components shown. The chat room agents 8, 9 are assigned to a chat room by the owner of a particular chat room or any other authorized person. As shown in Fig. 1, chat room agent 9 (CRA 2) logs on to chat room 7. Chat room agent 9 (CRA 2) has been launched by the owner which is illustrated by line 10 in Fig. 1. Chat room agent 8 which is launched by the chatter web server 1 is logged on to the second chat room 6.

The chat room agents 8, 9 are able to perform a number of tasks. One of the tasks consists in logging chat conversations. The chat room agents also are programmed to do a content analysis. This is done by using standard text analysis techniques like parsing, identifying parts of speech of words, categorization of these words etc. The chat room agents 8, 9 can do contribution analysis, including the number of sentences contributed by each participant in a chat conversation. Another of the tasks handled by the chat room agents 8, 9 is facilitating chat conversations by delivering warning or alert messages to the relevant chat participant, including warning messages to a participant, when he uses expletives or undesirable language. They also can send messages to improve the use of English when a participant makes grammatical mistakes.

The chat room agents 8, 9 are either a software program logged on to the chat room with a persona, like any other human user or are integrated to the chat room 5, 6, 7 under a system which has been published under the name KRDL-Chat. KRDL-Chat is a particular embodiment of a chat system integrated fully with chat room agents and personal chat agents.

It is important to note that the chatter web server 1 works with any chat system which incorporates or allows the use of chat room agents or personal chat agents.

The personal chat agents 2, 3, 4 act as single client for multiple protocols. They can perform logging (when requested), communicating with a chat room agent (if applicable) like CRA 2 and other personal chat agents to obtain information, do user profiling, eavesdropping and matchmaking (with the help of other PCAs and CRAs), filtering, recommendations of chat rooms and alerting.

The core of the chat room agents 8, 9 and the personal chat agents 2, 3, 4, is generic. Their application program interfaces (API) allow them to interface with different types of chat facilities.

As it is illustrated in Fig. 1, a chat room agent or a personal chat agent needs not to register with the chatter web server 1 when it is running. However, without registering with the chatter web server 1 , it will not be -able to talk to other agents but can only perform stand alone tasks.

When the network communication system is used by a user, he will start his personal chat agent 2, 3 or 4 on his personal computer. He then can select to register his personal chat agent 4 with the chatter web server 1 which is illustrated by line 12 in Fig. 1. His personal chat agent 4 also serves as chat client, which allows him to log on to one or more chat rooms 6, 7. The personal chat agent 4 of the user also allows him to request one of a plurality of services provided by micro-agents. If the user owns any chat room like chat room 7, he can run a chat room agent 9 logged on to the chat room 7 to do eavesdropping. If his personal chat agent 4 is registered with the chatter web server 1 , he may request through his personal chat agent 3 eavesdropping to be done on one or more chat rooms 5, 6, 7. The actual eavesdropping may be done by other users' personal chat agents 2, 3 and chat room agents 8.

The invention allows agent services to be made available to other types of online services. If there is an online service with a chat facility, this service can become a part of the network communication system according to the invention which is called chatter web by simply adding to the chat client component of the online service a routine to interface with the application program interface of a personal chat agent. Chat contents then will be "multiplexed" to both the chat window and the personal chat agent. There is a separate API for chat room agents to log on as a normal user, but which only participates in chat.

Each personal chat agent 2, 3, 4 or chat room agent 8, 9 contains plug-in-micro-agents, e. g. logger, content analyser, contribution analyser, eavesdropper etc. which is illustrated in Fig. 2. The features of micro agents shown in Fig. 2 include:

- Logging: maintains raw log, or clean up the log; send log files upon request;

- Content analysis: language filtering (instant replacement/elimination); topic identification; spam detection; quick summary;

- Contribution analysis: by user (e. g., proportions of participants' contributions within one chat session or one project - a group of chat sessions); by topic (e. g., proportions of topics being discussed in one or a group of sessions); other chat-related statistics (e. g. average time a chatter stays in a room);

- User modelling: (1) interests; (2) habits (patterns like "logging on to chat room A between 2pm-3pm on every Friday"); (3) reputation rating (usage of undesirable words; spamming; other undesirable behaviours);

- Room modelling: (1) discussion topics; (2) overall reputation rating; (3) patterns (e. g. active during fixed period of time weekly => generates "informal" chat schedules for users' reference); (4) quality rating - "on-topic index";

- Group modelling: pre-specified group or aggregating user profiles to form communities of like-minded groups;

- "Chatter web modelling": 2-D clustering of users (given a "big" topic with many sub-categories); 2-D clustering of rooms; automated generation of hierarchical categorizations of rooms; - "Targeted" eavesdropping: eavesdropper specifies a room or a user to eavesdrop => 2 modes: (1) displays full conversation; (2) alerts the eavesdropper under certain conditions.

Fig. 3 shows an embodiment of a user interface for the personal chat agent 2.

The network communication system according to the invention allows for matchmaking and producing of dynamic links. Three lists are generated by the chatter web server 1. One of the three lists comprises a list of users with the same interests which, however, is not shown in Fig. 3. A second list with relevant chat rooms or chat links is illustrated in Fig. 3 as well as the third list referring to relevant web pages for which a web spider or a data mining tool is needed.

A personal chat agent has two possibilities for displaying the lists to its owner. The first one is a spontaneous recommendation during a chat session which lists changes dynamically if a chat topic is switched and/or if better matches are found. The second possibility relates to "static" recommendations based on the user's interest.

For chat room recommendations, if a discussion is currently carried out in a chat room, the room will be highlighted to online users who have similar interests. It is also possible to include software to have a special warning of reputation rating or distasteful contents (if applicable) for each chat link, so that the user can avoid joining such a room and getting offended.

As it can be understood with reference to Fig. 3, "a web of chat rooms" is achieved by dynamic generation of linkages which is analogous to web hyperlinks of the World Wide Web. In a particular embodiment, chat traffic lights may be provided which annotate each chat link with coloured dots to indicate the speed of the network link to that particular server.

The chatter web server 1 of Fig. 1 also compiles a "Chatter's Calendar" with pre-specified chat sessions for different categories of interest for PCA users to refer to. This can be done based on either the announcements which chat room owners have sent to the chatter web server 1 or can be based on room modelling done by the chat room agents 8, 9 of Fig. 1.

Entries into the "Chatter's Calendar" can have different colours in order ro reflect

1) either the chat session is announced by the room owner or based on room modelling; 2) public or private;

3) eavesdropping is allowed or disallowed;

4) confidence rating of interest modelling.

According to a preferred embodiment and referring to Figures 4 and 12, the network communication system is provided with what can be called a "virtual-virtual chat room". Since any existing chat room channel on the Internet can be referred to as a "virtual chat room", the term "virtual-virtual chat room" is meant to describe a third chat room into which parts of basic chat rooms are mirrored. Virtual-virtual chat rooms (WCR) are essentially spontaneous chat rooms which can be created when the chatter web server 1 matchmakes two rooms that are discussing similar topics at the same time under the condition that there is at least one chat agent 8, 9 in each room 5, 6, 7. When the two rooms are private and therefore do not allow non- members to access them, it is necessary for the chatter web server that the creation of a virtual-virtual chat room that has been recommended is accepted by the users in both rooms. If this is the case, the chat participants in each room need not quit the original room in order to log on to the virtual-virtual chat room because an eavesdropping window will be popped up on each personal chat agent client. The agents in the chat rooms send the utterances in each chat room to the virtual-virtual chat room which are displayed on the eavesdropping window collectively as if both groups of the chat participants were in the same chat room. From the chatter web servers' point of view, this is a special way of eavesdropping which can be called "synchronous" eavesdropping. Hence this spontaneous chat room is a virtual chat room that "sits on top of a group of virtual chat rooms and therefore can be called a virtual-virtual chat room (WCR).

In order to illustrate the way a virtual-virtual chat room works, Fig. 4 shows a WCR window in the upper part of the drawing in which users in different chat rooms are represented by different fonts.

Fig. 4 also shows the chat windows of a first and second chat room channel which are used to create the virtual-virtual chat room. As it can be seen in Fig. 4, the users can type in either a WCR utterance which will show up in the virtual-virtual chat room or a "private" utterance which will not be mirrored into the WCR window. In Fig. 4, different kinds of fonts are used in order to illustrate where the utterances are keyed in and where the utterances can be read.

Fig. 12 shows a flow chart illustrating the room-to-room matchmaking and the creation of a virtual-virtual chat room by the chatter web server 1. The chatter web server 1 receives the results of the real-time content analysis from the personal chat agents 2, 3, 4 and the chat room agents 8, 9 located in various chat rooms 5, 6, 7. In a next step, the chatter web server 1 looks whether a match can be found between chat rooms that are discussing the same topic at the same time. If this is the case, a recommendation for creating a virtual-virtual chat room is sent through the connected agents to all chatters in the matched chat rooms. If all chatters in the matched chat rooms agree, a new virtual-virtual chat room is created and handled. This process is executed regularly until the system shuts down.

Fig. 5 illustrates the connections between personal chat rooms, chat room agents, chat rooms and the chatter web server. It is essential for personal chat agents and chat room agents to log on or to register with the chatter web server so that each agent can locate and talk to the other agents on the chatter web created by the network communication system according to the invention. When a personal chat agent or a chat room agent logs on to the chatter web server, the agent sends the necessary log-in information to the chatter web server and requests to be authorized. Once it passes the authentication, a TCP/IP connection between the agent and the chatter web server will be established for subsequent communications. When a personal chat room agent logs on to the chat room A, it uses the owner's persona. In a similar way, when the chat room agent logs on to chat room B, it logs on to the room with its own persona.

Fig. 7 illustrates the process of creating a virtual-virtual chat room. After a content analysis of the chats in the chat rooms and after matchmaking by the chatter web server, the chatter web server recommends to create a virtual-virtual chat room through the agent. If the chatters agree in both rooms, the WCR is created which is shown in Fig. 7 within the CWS. As illustrated, the agents in the chat rooms send their utterances to the WCR which in turn "multicasts" the utterances to the agents so that they can be read by the users.

The process of user-to-user /-room /-web site matchmaking is illustrated in Fig. 6. A number of chat rooms or personal chat agents send the profiles of chat rooms or profiles of users to the chatter web server. Also information from a web spider is inputted to the chatter web server to do the matchmaking. The chatter web server consolidates the information into a data base and recommends rooms, users and/or web sites. These recommendations 20 are transmitted to a personal chat agent 2 which has issued a request for recommendations 22 to the chatter web server 1.

Fig. 11 shows a flow chart of user profiling which is performed by the personal chat agents to 3, 4. After a start 30 has been initiated, the personal chat agent executes steps 31 to 35 shown in Fig. 11 before the end of the profiling is reached in step 36.

The following features for user profiling / modelling performed by the personal chat agents are worth noting.

1. Data structure of a user profile:

- Chat history: Room and time period of each chat session that the user was involved in. - Interests: A list of "topics", each associated with a confidence rating (1-100; 100 represents the highest confidence)

2. "Topics" are structured in hierarchical categories.

3. "Glossary": - A list of keywords that are associated to each topic.

- Each keyword is assigned a "relevance rating" (0.1-1) that represents how relevant it is to the topic. For example the keyword "Windows_98" under the topic "operating system" can be assigned a rating of 0.9; while the same keyword under the topic "Internet" can be assigned a rating of 0.2.

- A PCA can either download the full or partial glossary from the CWS and store it on local harddisk for offline references, or perform online reference to the glossary on CWS provided it is registered to the server. - Human "glossary builders" are in charge of maintaining the topic hierarchy, updating the glossary and deciding on "relevance rating" of each keyword-topic mapping.

4. Initially, the PCA asks the user to specify his interests from a list of higher-level topics (e. g., to specify "music" instead of "Wolfgang Amadeus Mozart's works" and "country music", or to specify "information technology" instead of "Java programming" and "USB technology"). This is to free the user from detailed specifications and to provide the PCA some convenient starting points of interest modelling.

Each topic being specified will be assigned the confidence rating of 25.

5. The PCA analyses the user's utterances, picks up keywords and looks for corresponding (relevant) topics. The associated rating of each keyword will be accumulated to the confidence rating (maximum: 100) of its corresponding topic in the user profile. If the same keyword is relevant to more than one topic, then the rating of all relevant topics in the user profile will be updated. 6. The keyword ratings accumulated to the topic ratings in he user profile can be degraded over the time. The scheme is:

- Keywords mentioned in the past 3 months: Rating x 1

- Keywords mentioned between 3 - 6 months before: Rating x 1/2 - Keywords mentioned between 6 - 9 months before:

Rating x 1/4

- Keywords mentioned between 9 - 12 months before: Rating x 1/8

- Keywords mentioned more than 12 months before: Rating elapsed

Thus, a user's (shifts of) interests can be traced automatically. 7. Heuristic for refinement of topic categorization: If a (high- level) topic (e. g., "music") has a confidence rating of less than 30 and it has at least one descendent topic (e. g., "Mozart's works" or "country music") that has a confidence rating of more than 75, then the (high-level) topic is eliminated.

- The motivation behind this heuristic: e. g., if a user loves "heavy metal" but hates "classical music", it is obviously inappropriate to take the more generalized topic "music" as his interest (this is to prevent uninterested matchmaking).

8. Profile editor can be provided for the user to view and refine/correct the profile.

9. User profiles are sent to CWS and kept in a data base for future matchmaking and recommendations.

Whereas the user profiling is performed by the personal chat agents, the room profiling or room modelling is performed by the chat room agents. The data structure of a room profile comprises:

- Room theme(s): Optional, specified by room owner or chat room agent owner.

- Room history: up and down time; times when chatters enter and leave the chat room, patterns as discussed below.

- A list of topics being discussed: Each topic is associated with a confidence rating between 1 and 100. The number 100 represents the highest possible confidence rating.

The chat room agents detect topics of interest of the chat rooms based on the same scheme that applies to the personal chat agent. The room profiling/modelling performed by the chat room agent also analyses "patterns" of room history: The chat room agent detects active sessions. Active sessions are those where at least three chatters log on to the room for at least 15 consecutive minutes. Both parameters can be configured by the owner of the chat room agent. In this way, regular log-in patterns can be induced like e. g. "The room is active between 3pm and 4pm on every Sunday",

Room profiles are sent to the chatter web server 1 and kept in a data base for future matchmaking and recommendations.

In a preferred embodiment, the network communication system comprises software which allows "spontaneous" content analyses. For that purpose, content analysers are provided in the personal chat agents and the chat room agents which are capable of performing a "spontaneous" content analysis in a room. This is done by analysing the topics which have been discussed in a particular room during the last 10 minutes or any other predetermined time which can be configured depending on the circumstances. The personal chat agents or the chat room agents constantly update the chatter web server 1 on the "current topic of the room". The chatter web server 1 consolidates the updates and stores them in the room data base which is illustrated in Fig. 6. The content analysis is useful for making recommendations and matchmaking.

A flow chart of user-to-user /-web page /-room matchmaking which is performed by the chatter web server 1 is shown in Fig. 10 and comprises the steps 40 to 46. The flow chart illustrates in more detail the process of matchmaking shown in Fig. 6.

The recommendation list consists of three sub-lists. These are (1) chatter web users, (2) web pages and (3) chat rooms. The basis of matchmaking can be defined as follows: Matching the user's interest list with

(1 ) for the sub-list of users - according to the interest list in the user's profile;

(2) for web pages - according to the keyword list for each web page (web pages are detected and analysed by the web spider of the chatter web server); (3) for chat rooms - according to the topic list in the room profile and "current discussion topic".

When a personal chat agent is connected to the chatter web server, the chatter web server can provide an online recommendation/matchmaking service upon request. The server will recommend the five best matches for each sub-list. The number of best matches is configurable by the owner of the personal chat agent.

Concerning the matchmaking process for the sub-lists (1) and (2), the following details are of interest: Ignore the last 20 percent of interests in the user's profile that have the lowest confidence rates and interests with confidence rates lower than 20.

The following example will illustrate in more detail the matchmaking process: After the filtering process, user A has 5 interests remained: (1) Java programming (confidence rating 90); (2) stock exchange (80); (3) classical music (60); (4) fishing (40); (5) wine (35).

- The confidence rating of each of the remaining interests (it is referred to as "major interests") becomes the latter's "weight" (e. g., the "weight" of user A's interest on Java programming is 90).

- The chatter web server searches for users / web pages with 5 highest "weighted match (WM)" as defined below:

WM = Σ (userjnterest * weight)

An example of calculating WM for another user B who has the following interests: (1 ) classical music (80); (2) wine (70); (3) stamp collecting (55); (4) web publishing (45); (5) Java programming (40); (6) fishing (15); (7) action films (10) -

Classical music: 60 ("weight") * 80 (user B's confidence rating)

Wine: 35 ^* 70

Java programming: 90 * 40

Fishing: 40 * 15

(No other overlapping interests available.)

Hence, WM = 60 * 80 + 35 ^* 70 + 90 * 40 + 40 ^* 15

The recommendation is dynamic. That is, the recommendation given by the chatter web server at different times may have different results, as the profiles of the users across the server are constantly changing, and new users and web pages may be added.

The matchmaking process for sub-list 3 is very similar to the process for sub-list (1) and (2) with the following exception: - Based on the spontaneous content analysis made by agents of the active rooms, any room that is currently discussing a topic of any major interest of an user will be prioritised in the recommended list for the user.

Now, the recommendations and matchmaking (room-to-room) / virtual-virtual chat room are explained in some detail: The matchmaking process is similar to that of user-to-room matchmaking. However, if both rooms are discussing the same topic at the same time, the matching will be prioritised (see Fig. 6).

Room-to-room matchmaking facilitates the recommendations of creating virtual-virtual chat rooms (WCR) (see Fig. 7). - Based on results of "spontaneous content analysis" made by connected agents, the chatter web server 1 matchmakes two or more rooms that are discussing the same topic at the same time.

- The chatter web server 1 sends the recommendation of creating a WCR through the connected agents to all the chatters in the existing rooms. The agent will "speak" like a normal user on behalf of the chatter web server 1. "There is another room [click here to learn more about this room] that is also discussing <topic> at the moment. Would you like CWS to create a WCR so that you can meet and talk to the chatters of the other room?"

- If chatters in both rooms agree, the chatter web server 1 will create a virtual-virtual chat room as has been explained earlier.

- The chatter web server 1 appoints one of the chat agents in each room as a "WCR co-ordinator", in case there are more than one chat agents in the same room. When an utterance is made in room A, its WCR co-ordinator will send a message to the WCR. The WCR then multicasts the utterance to the WCR co-ordinator(s) of room(s) B (C, D, ...). - The other co-ordinators then "repeat(s)" the utterance according to a special convention, e. g. "<userlD> in <room_name> says, <utterance>." This is to benefit those chatters in the same room who don't have a PCA - they can "listen" to and participate in the WCR discussion using their original chat client. For those chatters who are running a PCA, the PCA will launch a window to display the utterances from other room(s) so that the user interface would be more organized.

One of the interesting features of the network communication system is the targeted eavesdropping. The process of launching an eavesdropping service is illustrated in Fig. 8 and 9. Fig. 8 illustrates the sequence of (1) request for service by a personal chat agent, (2) request for service by the CWS and (3) the eavesdropping services. Fig. 9 illustrates the steps 50 to 56 involved in the process of launching an eavesdropping service. The following comments are useful in understanding the targeted eavesdropping.

1. Type (1): User specifies one or a group of chat rooms to eavesdrop

- The user's PCA sends a request to CWS.

- The CWS looks for agent(s) in the chat room(s) that can provide the eavesdropping service. - When the CWS identifies an agent (if there are more than one, CWS will pick the agent with the least or no workload), it will negotiate with the agent on behalf of the PCA that requests for the service. Whether the agent accepts the job would depend on its owner's spontaneous decision or his configuration setting of the

PCA's long-term service policy.

- If the agent agrees, an eavesdropping session will be created and the CWS will act as the "middleman" of the communications between the requestor and the service provider.

- During the course of eavesdropping, there will be certain ways to alert the chatters in the room being eavesdropped. For example, the agent that provides the service will alert the chatters periodically by an utterance, "PCA says, <userlD> is eavesdropping on this room right now. [Click here to learn more about <userlD>.]"

- Different types of eavesdropping:

The service provider sends all the utterances and activities in the room to the requestor. The requestor (PCA) will launch a window to display these to the user.

The service provider alerts the requestor under certain conditions that are pre-specified by the requestor, e. g., when a specific chatter or a group of chatters enters the room; a keyword is mentioned for several times; a topic is being discussed (based on "spontaneous content analysis"); the requestor's name is mentioned; swear words (this is useful for teachers to monitor students' chat sessions). 2. Type (2): User specifies another user to eavesdrop (e. g., to eavesdrop an expert's "speeches" in one or various rooms)

- The user's PCA sends a request to the CWS.

- The CWS negotiates with the PCA of the "eavesdroppee" if the latter is connected and registered to the CWS.

- During the course of eavesdropping, there will be certain ways to alert the chatters in the room where the "eavesdroppee" is in. For example, the agent that provides the service will alert the chatters in words, "<userlD> is eavesdropping on this room right now.

[Click here to learn more about <userlD>.]"

- The requestor's PCA launches a window with one or several frames - each displays the utterances in one room where the "eavesdroppee" is in (the "eavesdroppee" may be in more than one room at the same time, hence, several frames).

As it can be understood from the above discussion, the described invention comprises a software system for generating dynamic links between chat sessions, and other concurrent chat sessions and relevant WWW resources and content. Links are generated based real-time analysis of online chat content to identify the topic(s) of discussion. For an individual user, chatter web will also generate links to chat rooms or ongoing chat sessions that had or have conversations on topics relevant to the user's profile.

The software system allows any chat room or service to provide the capabilities. The system allows an online chat service provider to provide enhanced functionalities to their users and customers.

Participants of an ongoing chat session will receive recommendations (as links to) on:

1. other chat rooms with ongoing conversations on similar or related topics; 2. web and other resources and content which are about topics similar or related to the topic of the ongoing chat discussion.

An individual user who is interested in participating in an online chat on topics as reflected in his profile or as specified by him, can request and receive recommendations (as links to) on: 1. chat rooms or channels in which the ongoing chat conversations are about topics that match his profile or interests; 2. chat rooms in which previous chat conversation have been about topics that match his profile or interests, and the day of the week and the time of the day when such conversations occurred.

For an individual user, the system also provides ways of:

1. monitoring what is being conversed in pre-specified chat rooms or

2. monitoring what pre-specified chat participants converse about ("targeted eavesdropping") or 3. monitoring what is being conversed about when pre- specified words or phrases are being spoken about;

4. eavesdropping in a virtual-virtual chat room what is being said in one or more other chat rooms ("synchronous eavesdropping")

The system uses standard text analysis methods for analysing chat content to identify topics of discussion.

Claims

1. A network communication system wherein a number of user stations (2, 3, 4) are connected via a data network to a remote chat server hosting chat room channels (5, 6, 7) where text messages may be exchanged, comprising software with a multi-agent structure that is able to sample the chat content of all channels and in view of matching interests makes recommendations to the individual users of the system concerning links to other users, chat channels and resources on the network by parsing the chat content and by making use of a keyword-based model of interest, characterised in that the system is provided with a global chatter web server (1) that includes a generator for generating dynamic links between the individual user's stored profile or, on the basis of the real-time analysis of the user's chat, between the user's ongoing chat session and the other chat room channels in which related topics are treated in ongoing chat conversations or were discussed in previous chat conversations, independently of the specific system where the chat room channels are hosted, so that chat room channels from different chat servers in different parts of the world can be linked without a need for the user to log onto a new chat room channel to share part of their conversation.

2. A communication system according to claim 1 in which the link generator provides the user with recommendations on web and other resources in the world which are about topics similar or related to the topic of an ongoing chat discussion.

3. A communication system according to claim 1 in which the user's chat is sampled to build dynamically in real-time a profile for the user and / or a profile for a chat room channel.

4. A communication system according to claim 1 which provides synchronous eavesdropping of what is being conversed in one or more chat room channels by creating a virtual-virtual chat room.

5. A communication system according to claim 1 which monitors what is being conversed in pre-specified chat rooms or what pre-specified chat participants converse about.