GB2444519A

GB2444519A - Recommendation system

Info

Publication number: GB2444519A
Application number: GB0624550A
Authority: GB
Inventors: Makram Bouzid; Jean Millerat; Julien Perez
Original assignee: Motorola Inc
Current assignee: Motorola Solutions Inc
Priority date: 2006-12-08
Filing date: 2006-12-08
Publication date: 2008-06-11
Also published as: GB0624550D0

Abstract

A recommendation system comprises a selection processor 111 which selects items such as content items (e.g. television programmes, music etc). The recommendation system sends query messages to a first user indicating suggested content items and receives feedback message with a preference indication for the suggested content items from the user. The user interaction is achieved using a messaging service and may specifically be an instant messaging service. The recommendation system may specifically emulate a text based human chat dialogue. A learning processor 105 modifies a user preference profile for the first user by inferring preference values for item characteristics in response to the preference indications and at least one characteristic of the suggested items. A recommendation processor 101 generates an item recommendation in response to the user preference profile.

Description

RECOMMENDATION SYSTEM AND METHOD OF OPERATION THEREFOR

Field of the invention

The invention relates to item recommendation and in particular, but not exclusively, to content item recommendation for content items such as music clips or television and radio programmes.

Background of the Invention

In recent years, the availability and provision of multimedia and entertainment content has increased substantially. For example, the number of available television and radio channels has grown considerably and the popularity of the Internet has provided new content distribution means. Consequently, users are increasingly provided with a plethora of different types of content from different sources. In order to identify and select the desired content, the user must typically process large amounts of information which can be very cumbersome and impractical.

Similarly, an increasing number of services and applications with many different options and customisation features are becoming available to the user.

Accordingly, significant resources have been invested in research into techniques and algorithms that may provide an improved user experience and assist a user in identifying and selecting content, personalizing services etc. For example, Digital Video Recorders (DVR5) or Personal Video Recorders (PVRs) which comprise functionality for providing recommendations of television programs to the user based on user preferences are becoming increasingly popular.

More specifically, such devices can comprise functionality for monitoring the viewing/recording preferences of a user.

These preferences can be stored in a user preference profile which subsequently can be used to autonomously select and recommend suitable television programs for viewing or recording. E.g. a DVR may automatically record programs which are then recommended to the user, for example by inclusion of the automatically recorded programs in a listing of all the programs recorded by the DVR.

In order to enhance the user experience, it is advantageous to personalise the recommendations to the individual user as much as is possible. In this context, a recommendation consists in predicting how much a user may like a particular content item and recommending it if it is considered of sufficient interest. The process of generating recommendations requires that user preferences have been captured so that they can be used as input data by the prediction algorithm.

Typically, the user preference profiles are generated by the recommendation system monitoring the behaviour of the user and specifically monitoring which content is explicitly selected by the user. For example, a DVR monitors the recorded or real time television programmes which are watched by the user and uses this information to generate the user preference profile. However, although such a system is user friendly it results in limited information that can be used for the user interfaces as only current behaviour is monitored.

Alternatively, some recommendation systems explicitly request user's to define their preferences. Such systems tend to use dedicated user interfaces to interact with users in order to collect the user preferences and provide them with personalised recommendations. These user interfaces are typically either dedicated web pages or particular application graphical interfaces. Such interfaces requires a dedicated access to the application and furthermore requires the developers to design specific user interfaces adapted for the specific device used.

Also, such systems tend to require users to explicitly define preferences for different content categories or content characteristics which is inconvenient and difficult for a user who may not fully understand the purpose and functions of the content recommendation system.

Hence, an improved content recommendation system would be advantageous and in particular a system allowing increased flexibility, improved user friendliness, facilitated development, improved user interaction, improved preference information and/or improved performance would be advantageous.

Summary of the Invention

Accordingly, the Invention seeks to preferably mitigate, alleviate or eliminate one or more of the above mentioned disadvantages singly or in any combination.

According to a first aspect of the invention there is provided a recommendation system comprising: selection means for selecting a first item; means for sending a query message to a first user using a messaging service, the query message comprising an indication of the first item; means for receiving a feedback message comprising a user feedback from the user via the messaging service, the user feedback comprising a preference indication for the first item; means for modifying a user preference profile for the first user by inferring a preference value for item characteristics in response to the preference indication and at least one characteristic of the first item; and recommendation means for generating an item recommendation in response to the user preference profile.

The invention may allow improved operation and/or an improved user experience for a recommendation system. The invention may specifically allow a recommendation system to generate user preference profiles and recommendations based on practical, flexible, mobile, familiar and/or convenient user interactions. The invention may provide an entirely new approach for managing, altering and/or generating user preference profiles for content item recommendations.

For example, the user interactions may be performed with the user remotely located from the content recommendation system and/or without any means of presenting content items to the user. Also, the user interaction may be effected in a user friendly manner which is well-known to the user and e.g. without requiring any knowledge or understanding of the workings of the recommendation system or the requirements or desires for the user preference profile. For example, a user may be guided to provide information that is used by the content recommendation system by an interactive human-like communication between the recommendation system and the user. The user e.g. may not need to specify content item preferences for categories or characteristics of content items. Rather the content recommendation system may build a user preference profile based on simple answers by the user to various questions and suggestions generated and posed by the recommendation system in a way resembling a communication with another person.

The invention may in many embodiments allow a ubiquitous access to the content recommendation system using a standardised and familiar interaction. The development of interface means for the content recommendation system may be substantially facilitated and the same interface may e.g. be used with many different types of user devices. In comparison to many other recommendation systems, the invention may provide improved user preference information and thus potentially improved recommendations.

The recommendation system may e.g. be arranged to generate recommendations for customisation of services or for recommendation of content items, such as music, television programmes, adverts etc The messaging service may be independent from a content distribution system providing the content items. For example, the messaging service may be a communication service incapable of distributing the content items, e.g. the messaging service may have data rate and/or delay characteristics making it unsuitable for communication of high data rate and/or delay sensitive content items such as video and/or audio data streams.

The process of querying the user and receiving feedback may be repeated for a plurality of content items and the user preference profile may be generated by inferring preferences from the user feedback for the plurality of content items.

The messaging service may specifically be a text messaging service and/or may be a messaging service of a cellular communication system such as the Short Messaging Service (SMS) or Multimedia Messaging Service (MMS) . The messaging service may specifically be an instant messaging service and may specifically be a chat service.

According to another aspect of the invention, there is provided a method of operation for a recommendation system comprising: selecting a first item; sending a query message to a first user using a messaging service, the query message comprising an indication of the first item; receiving a feedback message comprising a user feedback from the user via the messaging service, the user feedback comprising a preference indication for the first item; modifying a user preference profile for the first user by inferring a preference value for item characteristics in response to the preference indication and at least one characteristic of the first item; and generating an item recommendation in response to the user preference profile.

These and other aspects, features and advantages of the invention will be apparent from and elucidated with reference to the embodiment(s) described hereinafter.

Brief Description of the Drawings

Embodiments of the invention will be described, by way of example only, with reference to the drawings, in which FIG. 1 illustrates an example of a recommendation system in accordance with some embodiments of the invention; and FIG. 2 illustrates a method of operation for a recommendation system in accordance with some embodiments of the invention.

Detailed Description of Some Embodiments of the Invention The following description focuses on embodiments of the invention applicable to a content item recommendation system and in particular to a content item recommendation system for recommending content items such as adverts, music or television programmes. However, it will be appreciated that the invention is not limited to this application but may be applied to many other recommendations such as for example recommendations for specific personalised characteristics of a service provided to a user thereby allowing a user customisation of the service.

FIG. 1 illustrates an example of a content recommendation system in accordance with some embodiments of the invention.

In the system of FIG. 1 a user messaging service is used to provide a user interface to the recommendation system. The messaging service is used by the recommendation system to provide a familiar, user friendly and potentially pseudo-human user interaction that generates user preference information for specific suggested items. Thus generation may be completely separate from any content distribution or presentation and may for example be generated when the user is away from the device used for e.g. viewing the content items.

A preference learning algorithm is applied to the data obtained from the user interaction to generate a user preference profile that can be used to generate item recommendations for the user. The learning algorithm uses non-trivial personalization technology such as, for instance, machine learning or data mining techniques to generate a non-trivial user preference profile such as, for instance, a probabilistic representation of preferences or a set of logic rules describing conditional preferences.

Specifically, the content recommendation system of FIG. 1 comprises recommendation processor 101 which generates recommendations of content items based on user preferences for individual users. The recommendation processor 101 is coupled to a user preference profile store 103 in which user preference profiles are stored for a number of users supported by the content recommendation system.

The recommendation processor 101 matches e.g. item information from item store 113 with collected preferences for a particular user in the profile store 103 to generate a recommendation list of items that closely matches the preferences. For instance, if logic rules are used to describe user preferences, the match test could be a simple rule condition satisfaction when applied to item information. If probabilistic representation of preferences is used, the match test may be that the result of the probability formula applied to item information exceeds a certain threshold. Of course, some other item filtering techniques could also be applied to generate the list of recommendations. For instance, since the profile store 103 contains user preference profiles for a number of users, collaborative filtering techniques based on similarities between users or items could be used.

Thus, in order to generate an accurate recommendation for a user it is important that the user's preferences are accurately captured and reflected in the user preference profile for that user.

In the content recommendation system of FIG. 1, a learning processor 105 is coupled to the user preference profile store 103 and is arranged to generate and modify a user preference profile for a user in response to preference feedback information provided by a user for specific content items. However, in contrast to conventional systems wherein a dedicated user interface is employed or user preferences are estimated based on the behaviour of the user when being presented with content items, the current system obtains the user preference indications by engaging in a messaging interaction with the user. This interaction may be completely independent of the presentation and distribution of the content items themselves.

Specifically, the learning processor 105 is coupled to a message processor 107 which is arranged to generate, transmit and receive messages using the messaging service.

Specifically, the messaging service may be a chat service based on textual communications between two people. In this case, the message processor 107 may emulate a person using the chat service and engaging in a dialogue with the user in order to retrieve user preference information for content items.

The message processor 107 engages in a dialogue by continuously and repeatedly generating suitable chat text messages and sending them to the user as well as receiving response feedback messages from the user. The message processor 107 is coupled to a message service interface 109 which is arranged to transmit and receive the messages in accordance with the specifications and protocols of the messaging service.

The messaging service may specifically be a messaging service which is aimed at low data rate communications between users. It may be entirely independent of any means for distributing actual content items and may specifically be incapable of or unsuitable for communicating the actual content items.

In the content recommendation system of FIG. 1, user preference information is achieved by repeatedly sending query messages to the user indicating a suggested content item. In response, the user may transmit a message indicating a preference value for this suggested content item. The message processor 107 may then proceed to suggest another content item in a new query message and may in response receive a new feedback message for this content item.

In the example of FIG. 1 a selection processor 111 is coupled to the user preference profile store 103 and the message processor 107 as well as to an item store 113. The item store 113 comprises information of a number of content items that may be suggested to the user to provide feedback on. It will be appreciated that the item store 113 may typically not store the actual content items themselves but rather may only store data characterising the content items.

Specifically, the item store 113 can for a television programme content item store meta-data indicating a transmission time, a duration, a genre, a title etc for the television programme.

The selection processor 111 is arranged to select content items to be suggested to the user. In a simple embodiment, the selection processor 111 may simply select content items randomly and/or may sequentially select content items in response to user feedback about previously selected and suggested items. In other embodiments, the suggested content items may be selected in response to the user preference profile.

As a specific example, the message processor 107 may initially request a content item to suggest to the user from the selection processor 111. The selection processor 111 accordingly selects a content item from the item store 113 and feeds this to the message processor 107. In response, the message processor 107 generates a text message describing the selected content item and transmits this to the user as a simple chat message. The chat message is specifically a text message written such that it is understandable to a user. For example, a text message such as "Please indicate your preference for the television programme titled "The Nine O'clock News (Rating Scale: 1-5)" may be generated and sent to the user.

When the feedback message is received from the user, the message is interpreted by the message processor 107 for example using a simple text parsing algorithm. The indicated user preference is then extracted and fed to the learning processor 105. The learning processor 105 is furthermore fed all the information stored in the item store for the content item. Thus, the learning processor 105 is in this way provided with a user preference value for a specific suggested content item as well as all characteristics known for that content item. These characteristics may for example include a category or genre for the content item, a director or artist associated with the content item etc. The process may be repeated a number of times (e.g. until the user stops providing feedback messages) and the learning processor 105 may proceed to generate/modify (the terms are used interchangeably) the user preference profile for the specific user based on the received preference indications and characteristics of a plurality of content items.

It will be appreciated that the recommendation processor 101 may use any suitable method for generating recommendations in response to the user preference profiles. For example, the recommendation processor 101 may have a number of predetermined content categories defined (for example corresponding to movies, sports, soaps etc) . The learning processor 105 may then infer a rating for each of these categories based on the user ratings! preferences received for the content items suggested to the user.

The recommendation processor 101 may then generate recommendations by receiving information about content items from item store 113 and matching them to user preferences from profile store 103 to select a number of content items that are eligible for being recommended to the user. The rating of each of these content items may be found by determining if they belong to any of the predetermined categories and if so assigning them the rating of this category. The highest rated content item(s) may then be recommended.

The learning processor 105 may use e.g. some clustering techniques, like K-means algorithm, to group user rated items based on similarities between their associated information, e.g. genre, and generate clusters of user preferences to store into user preference profile store 103.

The recommendation processor 101 may use e.g. a similarity measure to identify the cluster of preferences in a user profile which a new item should belong to, and a prediction function e.g. a probabilistic model to determine the rating of the new item based on the ratings of cluster items.

In the specific example, the recommendation system is arranged to use an instant messaging service which e.g. may be supported by a cellular communication system or the Internet. For example, the message service interface 109 may be arranged to interact with a mobile device to transmit messages in accordance with the specifications of a cellular communication system. Alternatively or additionally, the message service interface 109 may comprise Internet access functionality and an instant messaging client application compatible with an existing instant messaging chat service such as MSNTM messenger, YahooTM messenger or GoogleTM talk.

Instant messaging is a form of real-time communication between two or more users/clients based on exchange of messages. Specifically, instant messaging can be based on a repeated exchange of text messages with short delays thereby enabling an efficient real time text dialogue. The text is typically conveyed via computers connected over a network such as the Internet. Instant messaging typically includes the involved entities executing a client program that hooks up to a centralized instant messaging service and differs from e.g. e-mail in that conversations are then able to happen in real-time.

In early instant messaging programs, each letter appeared as it was typed, and when letters were deleted to correct typos this was also seen in real time. This made it more like a telephone conversation than exchanging letters. In modern instant messaging programs, the other party in the conversation generally only sees each line of text right after a new line is started. Most instant messaging applications also include the ability to set a status message.

Mobile instant messaging is a presence enabled messaging service that aims to transpose the desktop messaging experience to the usage scenario of being on the move.

Mobile instant messaging may in particular be supported in a cellular communication system thereby allowing the user interaction of the recommendation system to be achieved anywhere that cellular communications are supported.

In an instant messaging system, each user is identified by a unique identity that can be exchanged between users to enable them to contact each other for e.g. a chat. In the example, an instant messaging identifier is attributed to the recommendation system which accordingly appears as any other instant messaging client to the instant messaging system. Thus, in the example, the recommendation system is arranged to communicate via the instant messaging service using the messaging identity allocated to the recommendation system. The communication with the recommendation system is treated as any other instant messaging communication by the instant messaging service and indeed the recommendation system may to the instant messaging system be indistinguishable from a normal user.

In the recommendation system of FIG. 1, each individual user is associated with their unique instant messaging identity.

Specifically, each user is identified by the nickname provided by the instant messaging system. Thus, the learning processor 105 may determine that the user preference profile for a specific user should be evaluated/updated/generated.

Accordingly, it may instruct the selection processor 111 to select a number of content items to suggest to the user and may control the message processor 107 to initiate the user interaction to obtain preference feedback from these content items. In response, the message processor 107 then proceeds to execute a real-time instant messaging chat communication with the user.

In a specific example, the message processor 107 provides the message service interface 109 with the instant messaging nickname of the user as well as the first query message in the form of a text string. The first text string includes an identification of the first suggested content item in a form which is directly understandable by the user. The message service interface 109 proceeds to set up an instant messaging chat session with the identified user using standard instant messaging protocols and procedures.

At some stage, the user responds by transmitting a feedback message in the form of a text string including a user rating of the suggested content item. When the text string is received from the user by the message service interface 109 it is fed to the message processor 107 which passes the received text string to extract the indicated preference value. This preference value is fed to the learning processor 105. It then proceeds to generate the next text string with an indication of another suggested content item.

This text string is then fed to the message service interface 109 for transmission to the user. Thus, the message processor 107 can effectively implement an artificial conversation system for the dialogue with the user This approach is repeated until all selected content items have been suggested to the user or until the user terminates the session e.g. by entering a predetermined term in a response message. The learning processor 105 then proceeds to evaluate the received user preferences and content item characteristics in order to update the user preference profile. For example, if a high preference value has been received for a specific content item, the preference value for a category of the user preference profile to which the content item belongs is increased by a predetermined value.

Similarly, if a low preference value is received for a given content item, the stored preference value for categories to which this content item belongs is reduced.

Thus, using a, possibly existing, instant messaging infrastructure, the recommendation system is able to interact with any user of a compatible instant messaging program to learn their content/service preferences. For instance, it may propose some content items for rating and use the user feedback to model and learn their preferences.

The system can then use the determined user preference profiles to provide users with content recommendations that are specifically targeted to the individual user. For example, the recommendation system may recommend specific adverts to a content provider for distribution to the individual user.

In a centralised approach, the collected user preferences may enable the recommendation system to use various recommendation algorithms such as content-based filtering or collaborative filtering correlating preferences of several users, or a mixture of both.

In some embodiments, the selection processor 111 may be arranged to select the content items which are suggested to the user based on the user preference profile. For example, in some embodiments a large number of content items having characteristics which have a high preference value in the user preference profile may be suggested to the user. This will allow an increased amount of data to be generated for content items which are likely to have high preference values thereby allowing a more accurate and detailed evaluation and recommendation of content items likely to be of high interest to the user. Thus, a reduced effort is used to evaluate and obtain preference information for content items which are likely to be of little interest to the user.

This may provide a more accurate and detailed user preference profile.

In the system of FIG. 1, each user preference profile is linked to the messaging identity of the corresponding user.

This may facilitate the operation of obtaining feedback preference information and may also allow user recommendations to be determined based simply on the instant messaging identity. Thus, a user targeted content item recommendation may be provided simply in response to receiving an instant messaging identity.

Thus, any device having an instant messaging client may not only efficiently provide user feedback information but can also receive recommendations simply by requesting such a recommendation as part of an instant messaging session.

For example, as part of the session set up by the recommendation system in order to update a user preference profile or as part of a dedicated instant messaging session set up by the user, the user may provide an item recommendation request to the recommendation system. For example, the user may simply type a pre-defined term, such as "Recommend", during a chat session with the recommendation system.

In response, the recommendation processor 101 may be initiated to provide a content item recommendation for the user having the instant messaging identity of the session. The recommendation processor 101 thus retrieves the appropriate user

preference profile and the available items, and generates the recommendation. The recommendation is then fed to the message processor 107 which proceeds to generate a message comprising an identification of the recommended content item. For example, the message processor 107 may simply include the title of a specific content item in a text string sent to the user.

It will be appreciated, that the recommendation need not be provided directly to the user or to an instant messaging client operated by the user. For example, the recommendation may be provided to a content distribution server which may proceed to provide the recommended content item to the user via other means. In such examples, the interface between the remote entity and the recommendation system may still use the instant messaging service.

In some examples, the recommendation system may generate an indication of a user group comprising users with interest in a specific content item or group of content items. For example, a content distribution server may use an instant messaging service to transmit an indication of a specific content item to the recommendation system. The recommendation processor 101 may evaluate the user preference profiles to determine a group of users having a high preference value for the characteristics of the specific content item. The recommendation system may then transmit an indication of this user group back to the content distribution server which may proceed to distribute the specific content item to the selected user group.

In some embodiments, the recommendation system may be arranged to take the user context into consideration.

Specifically, the user context may be evaluated when generating the user preference profile and/or the user context may be used when generating a specific recommendation.

Specifically, in some embodiments the query messages sent to the user via the instant messaging service may not only request a preference value but also a current user context, such as the location of the user when providing the preference. The user may in response indicate the preference value as well as the current context. The context may be a simple selection between predetermined context such as a selection of one characteristic from the set comprising "Home", "Work", "Car", "Other". The user preference profile may accordingly be enhanced to also include context specific preferences.

Similarly, when a user requests a recommendation, the recommendation request may indicate a current context, e.g. by selecting a characteristic from the same set. The recommendation processor 101 may use this to retrieve the appropriate context specific user preferences when generating the recommendation.

Thus, the recommendation system may use information of the user context obtained by the messaging service to provide improve recommendations. In the example, the user context data is comprised in the message from the user.

Alternatively or additionally, the user context data may be data which is not explicitly provided by the user but is generated by and obtained from the instant messaging service itself.

Specifically, most instant messaging services provide a presence information feature which indicates whether users are currently online and available to chat. Based on presence information, the recommendation system may be able to take the context of the user into account (current activity, location...) . The integration with the instant messaging system provides an immediate access to this kind of presence information without having to develop and deploy further software components to grab information about the user's context. Thus improved recommendation performance may be achieved.

In some embodiments, the recommendation system may be arranged to operate in different modes and may be arranged to switch between these modes in response to instant messaging data received from the user. Specifically, the recommendation system may be arranged to enter a learning mode when a learning command is received in an instant messaging message from the user. The learning command may be a simple predefined term such as "Learn".

When the system is in the learning mode, the recommendation system proceeds to obtain preference value information by repeatedly suggesting content items and receiving preference values using the instant messaging service. The recommendation system may continuously update the user preference profile during the learning mode or may for example update a user preference profile when exiting the learning mode.

In the following, an exemplary instant messaging dialogue between a recommendation system and a user is provided. In the example, the system learns user preferences for television programmes and is able to recommend television programmes to the user. User hello

Recommendation System Hi, I am a Recommendation Bot, I can recommend TV programmes for you to watch For more information, type "help" User help Recommendation System To be able to recommend you something, I need to know you a little bit better.

To begin a learning phase, type learn', to stop the learning phase, just type "stop".

Moreover, you can type "help" to know what to type at every moment -"learn" -> to begin learning phase -"recommend" -> to have a new recommendation -"list' -> to view all your preferences I already know -"help" -> to obtain contextual help User learn Recommendation System Ok, I'll provide you with a set of TV programmes and you will rate them in a scale of 1 to 5.

What do you think about: title: Un billet de train pour...

channel: arte duration: 30 mins date: Mon, 7:30 pm star rating: 3 genre. [serie] User Recommendation System Ok, Thank you What do you think about: title: i-Cuisine channel: itv duration: 15 mins date: Wed, 4:35 pm genre [magazinej User Recommendation System Ok, Thank you What do you think about: title: Les experts sub-title: Circuit mortel channel: tf 1 duration: 60 mins date: Sat, 09:05 pm actors: [Paul Guilfoyle, William L. Petersen] directors: [Oz Scott] star rating: 3 genre: [série] User stop Recommendation System Alright, as you wish User learn Recommendation System What do you think about: title: Drinngz channel: europe2 duration: 60 mins date: Sun, 4:20 pm genres: [musique] User Recommendation System Ok, Thank you What do you think about: title: Emission spéciale subtitle: Quelle defense pour Ia France? channel: parlement duration: 60 mins date: Fri, 10:30 pm genre [débat] User Recommendation System Ok, Thank you What do you think about: title: Les ATnu de Hokkaido channel: arte duration: 55 mins date: Thu, 10:05 pm star rating: 2 genres: [documentaire] User Recommendation System Ok, Thank you What do you think about: title: Météo channel: tmc duration: 5 mins date: Tue, 8:55 pm star rating: 2 genres: [information] User stop Recommendation System Airight, as you wish User recommend Recommendation System I can recommend you title: L'amour en heritage channel: m6 duration: 130 mins date. Mon, 9:40 pm actors: [Stefanie Powers] directors: [Douglas Hickox] star rating: 2 genres: [téléfilm. drame] You can give me a feedback if you want User Recommendation System Ok. I'll try to find a better programme next time User recommend Recommendation System I can recommend you title: Vélo club channel: france2 duration: 30 mins date: Sun. 10:10 am genres: [magazine] You can give me a feedback if you want User Recommendation System Great you can now program your DVR to record it on htp://www.mydvr.com/?userxxxxzzzz-im.com In some embodiments, the interaction between the user and the recommendation system may be instigated by the user setting up a chat session with the recommendation system.

However, alternatively or additionally, the recommendation system may initiate the interaction. For example, the message service interface 109 may generate a messaging alert notification which indicates that the recommendation system seeks to set up a chat service with the user in order to obtain preference value information. Specifically, such an alert notification may be generated when a content item is available for suggestion to the user.

The alert notification may be treated in the same way as any attempt by a user of the instant messaging service to set up a chat session with another user.

The previous examples focused on the user interaction based on an instant messaging service. However, it will be appreciated that the described principles may be applied to many other messaging services. For example, non real-time messaging services supported by a cellular communication system may be used to interface with the user. Specifically the message service interface 109 may be arranged to generate, transmit and receive text or multimedia messages of an SMS or MMS message service.

It will be appreciated that the described approach of a recommendation system that can interact with users via a messaging service can provide a number of advantages including for example one or more of the following.

Designers of a recommendation system do not have to adapt user interfaces to each used device but can access a large number of people and devices using a simple common system.

Also, the system does not require the client component or communication protocol to be specifically designed, modified or extended with additional features for the purpose of supporting dialogues with the recommendation system.

Furthermore, the attention of content consumers is limited and content distributors find it difficult to engage with users to a desired extent. For example, content consumers are typically out-of-reach when they are not in front of their presentation devices (e.g. television sets) . This approach provides a practically ubiquitous access and reduces or eliminates the limitation that users have to be located by the content presentation/receiving device (e.g. a television) . For example, a user may use a simple standard messaging service on his mobile phone to customise his DVR Moreover, the approach can allow content distributors to be more seamlessly integrated into the daily environment of the user: no new software is required, there is no new web interface to learn, there is just one additional contact to add to the user's instant messaging contact list etc. Specifically, instant messaging offers a unique benefit to a user of recommendation systems because of its non-invasive "push" model: it is always on and can deliver pro-active recommendations without requiring any action of the user such as having to visit a specific web page whenever you Thus, the approach may provide a user friendly, flexible, practical and high performance recommendation system.

FIG. 2 illustrates a method of operation for a recommendation system in accordance with some embodiments of the invention.

The method starts in step 201 wherein a first item is selected.

Step 201 is followed by step 203 wherein a query message is sent to a first user using a messaging service. The query message comprises an indication of the first item.

Step 203 is followed by step 205 wherein a feedback message comprising a user feedback from the user is received via the messaging service. The user feedback comprises a preference indication for the first item.

Step 205 is followed by step 207 wherein a user preference profile is modified for the first user by inferring a preference value for item characteristics in response to the preference indication and at least one characteristic of the first item.

Step 207 is followed by step 209 wherein an item recommendation is generated in response to the user preference profile.

It will be appreciated that the above description for clarity has described embodiments of the invention with reference to different functional units and processors.

S However, it will be apparent that any suitable distribution of functionality between different functional units or processors may be used without detracting from the invention. For example, functionality illustrated to be performed by separate processors or controllers may be performed by the same processor or controller. Hence, references to specific functional units are only to be seen as references to suitable means for providing the described functionality rather than indicative of a strict logical or physical structure or organization.

The invention can be implemented in any suitable form including hardware, software, firmware or any combination of these. The invention may optionally be implemented at least partly as computer software running on one or more data processors and/or digital signal processors. The elements and components of an embodiment of the invention may be physically, functionally and logically implemented in any suitable way. Indeed the functionality may be implemented in a single unit, in a plurality of units or as part of other functional units. As such, the invention may be implemented in a single unit or may be physically and functionally distributed between different units and processors.

Although the present invention has been described in connection with some embodiments, it is not intended to be limited to the specific form set forth herein. Rather, the scope of the present invention is limited only by the accompanying claims. Additionally, although a feature may appear to be described in connection with particular embodiments, one skilled in the art would recognize that various features of the described embodiments may be combined in accordance with the invention. In the claims, the term comprising does not exclude the presence of other elements or steps.

Furthermore, although individually listed, a plurality of means, elements or method steps may be implemented by e.g. a single unit or processor. Additionally, although individual features may be included in different claims, these may possibly be advantageously combined, and the inclusion in different claims does not imply that a combination of features is not feasible and/or advantageous. Also the inclusion of a feature in one category of claims does not imply a limitation to this category but rather indicates that the feature is equally applicable to other claim categories as appropriate. Furthermore, the order of features in the claims does not imply any specific order in which the features must be worked and in particular the order of individual steps in a method claim does not imply that the steps must be performed in this order. Rather, the steps may be performed in any suitable order.

Claims

1. A recommendation system comprising: selection means for selecting a first item; means for sending a query message to a first user using a messaging service, the query message comprising an indication of the first item; means for receiving a feedback message comprising a user feedback from the user via the messaging service, the user feedback comprising a preference indication for the first item; means for modifying a user preference profile for the first user by inferring a preference value for item characteristics in response to the preference indication and at least one characteristic of the first item; and recommendation means for generating an item recommendation in response to the user preference profile.

2. The recommendation system of claim 1. wherein the recommendation system is a content item recommendation system, the first item is a content item and the item recommendation is a recommendation of a content item.

3. The recommendation system of claim 1 wherein the selection means is arranged to select the first item in response to the user preference profile.

4. The recommendation system of claim 1 further comprising: means for receiving an item recommendation request from a remote entity via the messaging service; means for transmitting an indication of the item recommendation to the remote entity via the messaging service; and wherein the recommendation means is arranged to generate the item recommendation in response to the item recommendation request.

5. The recommendation system of claim 1 further comprising user context means for determining a user context in response to user context data received via the messaging service; and wherein the recommendation means is arranged to modify at least one of the user preference profile and the recommendation in response to the user context.

6. The recommendation system of claim 5 wherein the user context means is arranged to determine the user context from user context data comprised in a message from the first user, the user context data originating at a messaging device of the first user.

7. The recommendation system of claim 5 wherein the user context means is arranged to determine the user context from user context data generated by the messaging service.

8. The recommendation system of claim 1 further comprising means for storing a plurality of user preference profiles; means for receiving a user recommendation request from a remote server via the messaging service, the user recommendation request comprising an item characteristic of at least one item; means for generating a user group indication in response to the item characteristic and the plurality of user profiles; and -means for transmitting an indication of the user group to the remote server via the messaging service.

9. The recommendation system of claim 1 further comprising means for entering a learning mode in response to a detection of a learning command in a message from the first user, the recommendation system when in the learning mode being arranged to iterate transmitting query items to the first user via the messaging service and modifying the user preference profile in response to preference indications for the query items received from the first user via the messaging service.

10. The recommendation system of claim 1 further comprising means for transmitting a messaging alert notification indicating that the first recommendation is generated for the first user.

11. The recommendation system of claim 1 further comprising means for storing a plurality of user preference profiles, each user preference profile being linked to a messaging identity of a user associated with the user preference profile.

12. The recommendation system of claim 1 wherein at least one of the recommendation message and the feedback message is a text message.

13. The recommendation system of claim 1 wherein the messaging service is a messaging service of a cellular communication system.

14. The recommendation system of claim 1 wherein the messaging service is an instant messaging service.

15. The recommendation system of claim 1 wherein the instant messaging service is a chat service.

16. The recommendation system of claim 1 wherein the recommendation system has an associated messaging identity, and the recommendation system is arranged to communicate via the messaging service using the associated messaging identity.

17. A method of operation for a recommendation system comprising: selecting a first item; sending a query message to a first user using a messaging service, the query message comprising an indication of the first item; receiving a feedback message comprising a user feedback from the user via the messaging service, the user feedback comprising a preference indication for the first item; modifying a user preference profile for the first user by inferring a preference value for item characteristics in response to the preference indication and at least one characteristic of the first item; and generating an item recommendation in response to the user preference profile.