WO2009077901A1

WO2009077901A1 - Method and system for enabling conversation

Info

Publication number: WO2009077901A1
Application number: PCT/IB2008/055043
Authority: WO
Inventors: Peter Bingley; Berent W. Meerbeek; Albertus J. N. Van Breemen
Original assignee: Koninklijke Philips Electronics NV
Current assignee: Koninklijke Philips NV
Priority date: 2007-12-18
Filing date: 2008-12-02
Publication date: 2009-06-25
Anticipated expiration: 2010-06-18

Abstract

The present invention relates to a method of enabling a user (310) of a chat system (300) to engage a further user selected from a plurality of users in conversation as well as to a chat system (300) for providing chat functionality to a user (310), the chat system (300) arranged to enable the user (310) to engage a further user from a plurality of users in conversation, the chat system (300) comprising: means (315) for collecting user specific conversation context information indicative of the context of past conversations of the user with other users; means (325) for correlating user specific conversation context information from the user with user specific conversation context information of other users from the plurality of users; means (335) for selecting the further user from the other users based on the correlating; and means (345) for enabling the user to engage the further user in conversation.

Description

Method and system for enabling conversation

FIELD OF THE INVENTION

The present invention relates to a method and chat system for enabling a user to engage in conversation with other users.

BACKGROUND OF THE INVENTION

In the last decade a new generation of computer users has discovered that online conversation in the form of chat-systems does represent a simple and efficient means of communication. Chat clients such as MSN are more popular then ever and enable new communities to be formed on-line. Although on-line chat systems experience an unprecedented popularity, the underlying technology originates from several decades ago. Internet Relay Chat (IRC) and Bitnet Relay are examples of such systems that originate from the 1980s, that provided online conversation or conferencing at that time to those connected to the Internet or Bitnet.

However in spite of the wide-spread availability, it is often difficult to find proper candidates for conversation. To address this issue chat systems generally comprise multiple topical chat-rooms, virtual locations wherein multiple users can engage in conversation on a particular topic. Usually such conversation is text-based conversation, but more recently audio and/or video based conversation systems are being used.

Finding a proper conversation partner however remains an issue.

SUMMARY OF THE INVENTION

It is a goal of the present invention to provide a chat system that provides an alternative manner for enabling a user to engage in conversation.

This goal is achieved by means of a chat system for providing chat functionality to a user, the chat system arranged to enable the user to engage a further user from a plurality of users in conversation, the chat system comprising: means for collecting user specific conversation context information indicative of the context of past conversations of the user with other users; means for correlating user specific conversation context information from the user with user specific conversation context information of other users from the plurality of users; means for selecting the further user from the other users based on the correlating; and means for enabling the user to engage the further user in conversation.

The present invention proposes to enhance existing chat-systems that already allow a user to engage in conversation with further users in that it actively collects user specific conversation context information. This information is indicative of the context of past conversations of the user with other users. In this manner information such as whom did the user converse with, what were the topics discussed and when were these topics discussed can be collected.

This information in turn can be correlated with information from other users in order to establish e.g. whether both users have engaged in conversations on the same topic, with the same people and/or around the same time. The outcome of this correlation can subsequently be used to select a further user from the plurality of users in the chat-system that represents a match with the user. Subsequently the chat-system enables the user to engage the further user in conversation. The latter may be done inside the chat-system itself, e.g. by providing the user with a selectable icon that initiates a chat session between the user and the further user. However it will be clear to the skilled person that a wide variety of alternatives can be used to enable the user and the further user to engage in conversation. In this manner the present invention effectively provides an alternative manner for enabling a user to engage in conversation. In an embodiment the chat system further comprises means arranged to provide a conversational agent, more commonly referred to as a chat bot, or chatter bot and wherein the conversational agent is arranged to engage the user in conversation. In this manner the present invention enables the system to initiate conversation on topics that facilitate the collection of user specific conversation context information. In a preferred embodiment comprising a conversational agent, the conversational agent comprises output means in the form of an animatable object on a display or alternatively in the form of an animatable physical object. The output means being arranged to provide at least part of the conversational agent feedback by animating the (physical) object. In this manner a (physical) representation of the conversational agent can be created that provides further visual information to the user of the chat system. By using such an animatable object the conversational agent is brought closer to the personal sphere of the user and non-verbal feedback can be provided to enhance the user experience. In a preferred embodiment comprising a conversational agent, the conversational agent is further arranged to engage the user in conversation on an agent selected topic in order to collect further user specific conversation context information. By having the agent actively select a topic for conversation, the conversation with the user can be made more personal and entertaining, while in the process the acquisition of information needed by the system, e.g. to select a proper conversation partner, can be made more efficient.

The present invention may be used advantageously in a context wherein a user confides personal or privacy sensitive information to the conversational agent. In such a scenario the information collected by the conversational agent is preferably stored in a privacy preserving manner. Privacy preserving collaborative filtering techniques can be used to correlate information from multiple users without the need to disclose information. In this manner people with a common interest can be brought into contact.

Apart from the use of conversation context information the present invention may also use further information such as well established user profiles in order to enable people to contact kindred spirits. In a further embodiment of the present invention the user specific conversation context information is acquired from, or augmented by, conversation context information from communication means outside the chat system. Here e.g. information from E-mail clients, Internet browsers, and/or IP-phone devices, or switched network telephone devices can be acquired. The objective is further achieved by means of a method according to claim 12 and a computer program product according to claim 14.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other advantageous aspects of the invention will be described in more detail using the following Figures.

Fig. 1 shows a block diagram representing a method according to the present invention;

Fig. 2 shows a block diagram representing a further method according to the present invention; Fig. 3 shows a chat system according to the present invention;

Fig. 4A shows a chat client device comprising an animatable physical object according to the present invention;

Fig. 4B shows a preferred animatable physical object for use in the method according to the present invention; Fig. 5 shows a block diagram providing an overview of the information flow to/from a conversational agent for use in the present invention.

The Figures are not drawn to scale. Generally, identical components are denoted by the same reference numerals in the Figures.

DETAILED DESCRIPTION OF EMBODIMENTS

Chat-systems, also known as network-based synchronous conferencing systems, have been available for quite sometime. In particular network-based text conferencing systems were already well-known in the 1980s. Examples of such systems comprise e.g. Internet Relay Chat (IRC). More recently chat-services that allow synchronous audio conferencing and/or synchronous video conferencing have appeared.

Chat-systems generally allow communication over so called channels. Users can be associated with channels; often referred to as the users "joining channels", thereby allowing them to observe ("passive participation") and/or take part ("active participation") in conversations on the channel. In this respect a channel effectively reflects a conversation between two or more people, however by virtue of the nature of a chat system the people need not be in physical proximity.

Different types of channels are available private channels, wherein access to the channel is restrictive and controlled, and public channels which users can join freely. Channels in a chat-system can exist without users being associated with them.

This feature is quite often used to "channel" particular topical conversations and enable people to find other users that are interested in a particular topic more easily. Alternatively a user can create a channel, and provide it with a topical label on an ad hoc basis. Such ad hoc created channels are generally deleted when the last user disassociates ("leaves") a channel. Even in the early chat-systems agents, generally referred to as bots, were deployed by the chat-systems that were used to distribute information to new users or even moderate conversations on a channel, triggering on keywords and forcing abusive users to leave ("kick users"). At the same time users started creating script based bots that were e.g. used to mimic natural users by providing seemingly intelligent responses when presented with questions.

Eventually the Artificial Intelligence embodied in the scripts and/or code underlying these conversational agents increased, thereby enabling conversational agents to mimic behavior known from the "Eliza" computer program by Joseph Weizenbaum in 1966. The present aims to provide a method for enabling a user of a chat system, such as those described above, to enable the user to engage a further user from the chat- system users in conversation. With reference to Fig. 1, a method according to the present invention comprises a first step Sl for collecting user specific conversation context information indicative of the context of past conversations of the user with other users. The conversation context information may comprise information such as topical information derived from conversation in which the user participated. Acquiring such information may in fact require analysis of a conversation thread in order to establish the topic of the conversation. Topic extraction can be as simple as counting the occurrence of pre-defined keywords, but may alternatively also comprise more complex semantic analysis of conversation threads and/or natural language. An example of such a technique is latent semantic analysis, a technique often used to generate keywords and/or abstracts from documents.

Alternatively the conversation context information may comprise participant information derived from conversations in which the user participated. By effectively monitoring the conversational partner that a user engages with, it is possible to establish a social contact circle of a users, additional information may be gathered e.g. by combining the topical information with the individual users. In this manner it can be made apparent which topics are preferably discussed with which other users. Likewise the frequency, the extent, and or tone of the conversation can be extracted and annotated to the social circle of users.

More alternatively the conversational context information may also comprise information with regard to the date and time that such conversations took place. In this manner, suitable conversational partners may be suggested that have not been in touch for quite some time, but who have been participating in extensive/frequent conversations in the past.

In accordance with the method of the present invention, the user specific conversational context information can be correlated with user specific conversation context information of respective further users, reflected by step S2. By correlating information from the respective user specific conversation context information an assessment can be made with regard to overlap in interests and/or commonalities among the respective users.

Although it can be envisaged that the correlating in fact performs an exhaustive correlation over all users, alternatives are also envisaged. For instance by limiting the number of users for which information is correlated using a user profile, and/or using a black list and/or white list to block and/or enable selection of particular users and/or topics. The latter feature is of particular interest for enforcing parental guidance settings and in addition also reduces computational load.

The correlating may furthermore be weighted in a sense that information from the user profile is used to increase the weights associated with particular information in the correlating.

The outcome of the correlation in turn can be used to select a further user that may represent a suitable, or the most suitable, conversational partner. This is reflected in Fig. 1 by step S3.

Finally the method according to the present invention enables the user to engage the further user in conversation step S4. Although preferably conversation is initiated within the chat-system this is not mandatory. Considering that chat-system client software generally is available within the context of a platform capable of multiple forms of communication, it may also be possible to enable the user to engage the further user in conversation by other means such as by means of an E-mail client, an Internet Phone or telephone. In fact it may suffice to provide the user with contact information that enables the user to contact the other user by other means not bound to the computing platform running a chat-system client, e.g. by using a switched network telephone or sending a postcard.

In the above manner the present invention effectively enables the chat-system to extract information from conversations that is subsequently used to recommend a further user as a conversational partner. Although the above example illustrates the selection of a single conversational partner, it may be beneficial to provide the user with a selection of conversational partners that are selected based on the correlation results.

In the latter case, the user could be enabled to review the potential conversational partners, by presenting the user with a public user profile of each of the respective conversational partners. More alternatively a user could initiate a group conversation by inviting one or more people from the group of people to join in a multi-user conversation.

Although the concept of opening a conversation with an automatically selected conversational partner may at first appear to represent a threshold, the present invention effectively provides a pre-selection based on historical data and known behavior of users, thereby effectively lowering this threshold owing to the established overlap or interest.

One potential area for application of the present invention may e.g. be on-line dating sites, wherein users and/or subscribers of the site can freely engage in conversations using a chat-system, and wherein conversational history is used to select potential partners. A further application area of the present invention may be a chat-system addressing the need of a growing group of people that are hampered by physical or other limitations to engage in social activities. To them, the present invention may represent an alternative for getting into contact with people that have some commonality. Likewise the present invention might be used in a chat-system for on-line support groups in order to facilitate people facing similar problems to engage in conversation.

Fig. 2 illustrates a further method according to the present invention that builds on the method described above. In the method presented in Fig. 2 the step of collecting user specific conversation context information step Sl is complemented by a step SO providing a conversational agent. The conversational agent is arranged to engage the user in conversation. In this manner the conversational agent can obtain additional information with regard to conversational context information in a more direct manner than through analysis of the historic information.

Fig. 3 illustrates a chat-system 300 according to the present invention comprising a chat client device 320. The chat client device can be implemented as a personal computer, or a personal digital assistant or smartphone provided with the proper software. The chat client device 320 is coupled by means of a network 350 with a chat server 330. Apart from the chat client device further chat client devices 331, 332 can be coupled to the chat server by means of the network 350. The chat client device presented here comprises means 315 arranged to collect user specific conversation context information indicative of the context of past conversations of the user with other users; means 325 arranged to correlate user specific conversation context information from the user 310 with user specific conversation context information of other users; means 335 arranged to select the further user from the plurality of users based on the correlating; and means 345 for enabling the user to engage the further user in conversation.

Although in this particular embodiment the means 315, 325, 335 and 345 are all located in the client device, representative of a "thick" client device. These means may also be partitioned in a different manner, wherein the chat client device effectively only comprises a user interface that allows the user 310 to interact with the means 315, 325, 335 and 345 which are executed on a web-based server.

It will be clear to the skilled person that the actual partitioning can be chosen in dependence of the system requirements such as the capabilities of the chat client devices, allowing e.g. personal digital assistants and/or mobile phones to act as thin client devices and personal computers as thick client devices.

The network 350 effectively provides connectivity for the chat client devices and the chat server. As such the present invention can be advantageously used in chat-system that utilizes a local area network, an intranet, and/or the internet. Alternative networks such as UTMS, or other wireless network technology may be employed to equal advantage.

The chat-system 300 may be implemented in software on programmable computer platforms. Alternatively the invention can be implemented using dedicated chat hardware. It will be clear from the above that intermediate implementations are also envisaged.

With reference to Fig. 4A, a chat client device 320 employed in the present invention may further comprise visual output means 400 comprising an animatable object. The object could be an on screen avatar, which may be a user selectable character, and may e.g. be selected from a set of avatars available from the chat-system. The output means 400 being arranged to provide at least part of the conversational agent's user feedback through animation of the object.

More preferable the animatable object is a physical animatable object which "personifies" the conversational agent to the user. Fig. 4B shows a set of images 400 of such a physical animatable object in the form of the Philips iCat with different facial expressions. iCat is a user interface robot that can provide audio and visual feedback to a user. In particular iCat is capable of rendering facial expressions allowing mood adapted message delivery.

Fig. 5 presents an overview of the information flow in a chat client for use in a chat system according to the present invention, the chat client comprising a conversational agent 510 and a user interface 520. The user interface 520 could be a user- interface robot with a humanlike face (such as the above described iCat robot (FIG 4B) or a virtual character on a screen.

The conversational agent 510 shown here is responsible for generating sentences. Sentence generation can be done using known conversational agent technology such as the Artificial Intelligence Markup Language (AIML). To generate sentences the conversational agent shown here uses six types of information.

The first source of information is knowledge about the user's direct life, such as names, preferences (e.g. music, food, television shows, books, etc.), relatives (children, friends, parents, grand children) and more general facts (weather, sport, news). This knowledge is stored in a database called world & user model 530. While some pieces of knowledge are learned by "mining" information from conversations with the user, the world & user model 530 may also receive information from the chat-system, e.g. information regarding other users of the system over a network interface 540. The network interface 540 could be an interface to the Internet or, if the system is deployed in an elderly care home, from an Intranet.

The second source of information used by the conversational agent 510 is information about previous interactions between the system and the user. For instance, when there was already a conversation about the birthday of one of the user's grandchildren, then the conversational agent 510 can either refer to this chat, or decide not talk about it again. Knowledge acquired over time is stored in a database chat history 550.

The third source of information used by the conversational agent 510 is information about which topics are preferred by the user. This information is stored in a database called user profile 560. This database contains for every topic weight values that indicate how important that topic is for the user. Here a fractional value between 0.0 and 1.0 is used, 0.0 indicating "not of interest", 1.0 indicating "of high interest". During a conversation the weights in the user profile 560 can be updated with new profile information. For instance, the conversational agent 510 might ask a user directly whether he/she is interested in sports. The answer is then stored in the user profile 560 for future usage. The fourth source of information used by the conversational agent 510 is the input from the user that is received from the user interface 520. This includes for example questions that the user asks directly to the system (e.g. "When will the new government be installed?" or "Will the weather be good for walking?"), but also monitoring by the system of discussions that the user has with other people (so that these topics can be stored and used by the system at a later time to start a conversation).

The fifth source of information used by the conversational agent 510 is input about the user's social contacts (e.g. email, phone calls) from a user social history database 570. This includes the topics that the user discusses during social contacts, which is expressed in the socialness of each topic, a fractional value between 0.0 and 1.0, wherein 0.0 indicates "never discussed", 1.0 means "always discussed"). The user social history 570 database obtains its information for example from the user's email system and/or by monitoring internet phone conversations.

The sixth source of information used by the conversational agent 510 is the topic of interest that was selected by a topic selection module 580. This module here uses a weighting mechanism to select a topic for the conversational agent 510 that is of interest for the user. For example, in one embodiment the system has 5 topics ["child John", "weather", "dogs", "sports", "finance"]. For each topic the topic selection module 580 obtains the topic weight from the user profile 560 database, the topic socialness from the user social history 570, and the topic activation level from a topics database 590. At a particular moment in time the system could have the following topics, weights, socialness and activation level vector:

topics = ["child John", "weather", "dogs", "sports", "finance"] weights = [0.3, 0.5, 0.4, 0.8, 0.6] socialness = [1.0, 0.2, 0.5, 0.2, 0.1] activation level = [85, 15, 42, 95, -39]

selection = weights x socialness x act.level = [25.5, 1.5, 8.4, 15.2, -2.34]

The topic selection module 580 would calculate the selection vector and select the topic "child John" as the current chat topic having the highest selection value (25.5) in the selection vector. All topics that are of interest to a user are stored in the topics database 590. A topic in this database is a dynamic variable (the activation level), who's value changes over time. The value of a topic in this example ranges between 100% (did not talk about this topic for a long time) and -100% (did talk too long about a topic). A topic manager 600 can add and remove topics from the topics database 590. This might be decided based on information that is available in the world & user model 530 and the chat history 550. The activation level of a topic variable is calculated by the following formula:

p_J (t) = a_j - p_J (t - \) + (\ -a_j ) - ^f_{J l} (u_l (t)) (1)

where

P_j(t) is the activation level of topic 7 at time t, a, is the time constant of topicy,

Ui(t) is an input received from the world & user model 530 or the chat history 550 that might influence topicy, fi_j(.) is a non- linear influence function that defines how input u^t) influences the activation level p,(t) The input function U₁(T) might represent how often words related to a topic have been mentioned during an interaction dialogue during a particular period, or the time since the last time a topic was active. While formula (1) uses linear dynamics, also a nonlinear dynamics models may be used, such as

P_J (t) = g_J(p_J (t -l),∑f „&,(!))] (2)

where g_y(.) is a nonlinear function.

As an extension the topic selection module 580 could exploit knowledge from the world & user model 530 about temporary occasions such as a birthday or the illness of a friend, holidays, etc. when performing topic selection. Another extension is to couple the chat system of the user, hereafter user A, with the chat system of a close friend, hereafter user B. In this manner the network interface 540 may be used to provide further information to the system. When, using information form the user profile 560 and the user social history

580 of the chat client of user A, the chat client of user A detects that user A would like to be contacted by user B. Subsequently the appropriate socialness values in the user social history 570 of the chat client of user B can be adapted so that the chat system of user B will take actions to stimulate user B to contact user A. Another extension is that the chat clients of users can exchange information among each other in order to favor the selection of topics that are of mutual interest (e.g. by increasing the socialness in the user social history 580 for topics that have matching topic weights in the user profile 560 of the chat systems of users that are well acquainted with each other), thereby stimulating the users to contact each other about the topic. Another extension is to use information related to use browsing patterns or picture browsing in a personal photo-database as an input for topic selection of the chat system, provided that suitable metadata is provided. Based on this information the socialness values can be adapted in the user social history 570, to stimulate the user to contact persons that are related to the pictures (e.g. "Didn't you go on that holiday with John? Maybe you should call him some time").

It will be clear from the above that a large variety of information can be used in by the conversational agent 510 to establish a topic for conversation. It is envisaged that this information at least can comprise: topical information derived from conversations in which the user participated, participant information derived from conversations in which the user participated, date information derived from conversations in which the user participated and/or location information derived from conversations in which the user participated. In a system wherein there is no conversational agent 510 that engages in conversation with the user the chat client will have to mine the relevant information from conversations that the user conducts in a more passive mode. Although feasible the use of a conversational agent 510 will substantially improve effectiveness.

As the information exchanged in a chat system can be privacy sensitive and/or personal, in particular when a conversational agent is employed which acts as a confidant to the user, the information in the system will have to be protected. Although typically security can be improved by communicating information under encryption this would still allow privacy sensitive information to fall in the hands of malicious parties, as generally such information has to be decrypted before it may be compared.

However in a preferred embodiment privacy is substantially preserved under correlation using a method as disclosed in International Patent Application WO2005/015462 A1R6 by the same applicant. This document discloses matching of data for use in collaborative filtering under encryption.

Although as suggested above this procedure may be applied in a chat system wherein a chat client comprises a conversational agent, this is not mandatory. In fact the same procedure can be applied in a system without such conversational agent.

As indicated above the present invention is not restricted to text based conversation systems, the present invention may also be employed in natural speech systems. Speech recognition technology capable of understanding unconditioned speech is well known. Reference is made here to the paper "The Philips automatic train timetable information system" by H. Aust, et al, Published in Speech Communication, vol. 17, issue 3- 4, November 1995. This paper discloses a technique that allows a caller to talk in unrestricted, natural, and fluent speech, very much like he or she would communicate with a human operator, and is not given any instructions in advance. This very technology can be used to enable the present invention for audio as well as video based chat systems. It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design many alternative embodiments without departing from the scope of the appended claims.

In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. It will be clear that within the framework of the invention many variations are possible. It will be appreciated by persons skilled in the art that the present invention is not limited by what has been particularly shown and described hereinabove. The invention resides in each and every novel characteristic feature and each and every combination of characteristic features. Reference numerals in the claims do not limit their protective scope.

Use of the verb "to comprise" and its conjugations does not exclude the presence of elements other than those stated in the claims. Use of the article "a" or "an" preceding an element does not exclude the presence of a plurality of such elements.

Claims

CLAIMS:

1. Chat system (300) for providing chat functionality to a user (310), the chat system (300) arranged to enable the user (310) to engage a further user from a plurality of users in conversation, the chat system (300) comprising: - means (315) for collecting user specific conversation context information indicative of the context of past conversations of the user with other users; means (325) for correlating user specific conversation context information from the user with user specific conversation context information of other users from the plurality of users; - means (335) for selecting the further user from the other users based on the correlating; and means (345) for enabling the user to engage the further user in conversation.

2. Chat system (300) according to claim 1, the chat system further comprising means for providing a conversational agent and wherein the conversational agent is arranged to engage the user (310) in conversation.

3. Chat system (300) according to claim 2, wherein the means for providing a conversational agent comprise visual output means (400) comprising an animatable object, the visual output means (400) arranged to provide at least part of the conversational agent's user feedback through animation of the object.

4. Chat system (300) according to claim 2, wherein the means for providing a conversational agent further arranged to engage the user (310) in conversation on a topic selected by the conversational agent in order to collect further user specific conversation context information.

5. Chat system (300) according to claim 4, wherein the further user specific conversation context information acquired by the conversational agent is handled in a privacy preserving manner.

6. Chat system (300) according to claim 2 or 4, wherein the means (400) for providing a conversational agent comprises means for conversation topic selection based at least in part on collected user specific conversation context information of the user.

7. Chat system (300) according to claim 1 or 2, wherein the further user selection comprises using user profile information of the user (310) as a further selection criterion.

8. Chat system (300) according to claim 1, wherein the user specific conversation context information comprises at least one of: topical information derived from conversations in which the user participated; - participant information derived from conversations in which the user participated; date information derived from conversations in which the user participated; and location information derived from conversations in which the user participated.

9. Chat system (300) according to claim 1, wherein the means for enabling the user to engage conversation with the candidate user comprises at least one of: a selectable on-screen representation representing the further user enabling the user to attempt to engage a chat session with the further user; a selectable on-screen representation representing the further user enabling the user to engage an mail-client to address a message to the further user; and a representation providing contact details of the further user to the user.

10. Chat system (300) according to claim 1, wherein the user specific conversation context information is acquired from, or augmented by, user specific conversation context information from communication means outside the chat system.

11. Chat system (300) according to claim 1 , wherein the means for selecting the further user are further arranged to select an other further user from the plurality of users based on the correlating; and wherein the means for enabling the user to engage the further user are further arranged to enable the user to engage the other further user in the same conversation.

12. A method of enabling a user (310) of a chat system (300) to engage a further user selected from a plurality of users in conversation, the method comprising: - collecting (Sl) user specific conversation context information indicative of the context of past conversations of the user with other users; correlating (S2) user specific conversation context information from the user with user specific conversation context information of respective other users from the plurality of users; - selecting (S3) the further user from the other users based on the correlating; and enabling (S4) the user to engage the further user in conversation.

13. A method of enabling a user (310) of a chat system (300) to engage a further user selected from a plurality of users in conversation according to claim 12, the method further comprising: providing (SO) a conversational agent and wherein the conversational agent is arranged to engage the user in conversation.

14. A computer program product on a computer readable medium, the product comprising instruction for causing a programmable device to execute the method according to anyone of claim 12 or 13.