GB2448874A

GB2448874A - Context based media recommender

Info

Publication number: GB2448874A
Application number: GB0708274A
Authority: GB
Inventors: Richard Hull; Stuart Philip Stenton
Original assignee: Hewlett Packard Development Co LP
Current assignee: Hewlett Packard Development Co LP
Priority date: 2007-04-30
Filing date: 2007-04-30
Publication date: 2008-11-05
Also published as: GB0708274D0

Abstract

A system for recommending media content related to a visit away from a home location 200 by a user, obtains using a mobile device, visit context information including a pattern of locations, automatically without prompting the user for manual entry. A recommender 230 matches, after the visit, the collected visit context information, and media content available for consumption after the visit. Recommendations are presented to the user after the visit, according to the matching. The recommendations can be more appropriate by determining them from the visit context rather than merely a present location, or past preferences. Collecting the visit context automatically can be more convenient and can broaden the recommendations to encompass things that the user was not aware of.

Description

CONTEXT BASED MEDIA RECOMMENDER

Field of'the Invention

The invent ion relates to systems for recommending to a user, media content related to a visit away from a home location, and to corresponding methods of providing or using a recommender service.

Background

Ii is known to provide systems having software Lbr making recommendations to users.

Such recommendation engines typically make recommendations based on similar actions. There are many examples of recommender systems that operate within some sphere of media-related activity. Television programs might be recommended based on stored preferences or a stored prolile of previous viewing by that user. For example. the Tivo IM Personal Video Recorder (PVR) will automatically record television programs that it considers to be of interest to a viewer who has explicitly watched (or recorded) some other program. A web based shop can prompt a user with possible purchases which might be recommended from pm ious purchases by that user or by similar users. For example Amazon IM recommends books, DVDs, CDs etc based on a user's prior purchases.

Systems that are capable of detecting and acting upon a mobile user's context are also well known. It is known to provide a mediascape system in which a user carries a mobile device such as a Personal Digital Assistant (PDA) which can detect its location, and without user input, play media content previously programmed as being appropriate to the location. An example has been created by HP, known as the Mobile Bristol client, which will monitor its user's location and perform niedia actions, such as playing an audio file, accordingly.

Patent application W003009191 relates to the storage and management of access to profile information, in particular to personal profile inibmiation. It shows a computer system having personal profiles or personal data files to allow a user's interaction with a system to be personalised. In addition to storing personal information such as credit card details, telephone number, postal address and so on, a personal profile may include details of the user's personal preibrences. for example to allow personalisation _) -of a portal website with the user's specific interests. The prolile may include current and/or historic user location information, for example forming a "location trail". This information may then be used either by a profile server or by an application accessing the profile server to predict where the user is going to he at some future time.

US patent application 2006069749 shows a location enhanced information delivery system which presents the information most suited to the real current audience, as measured by location information systems, rather than to a static predicted audience.

Location information can be proided by a beaconing-style ireless technology, or license-plate scanning with cameras. Location-based metadata can be used to activate rules that filter information on behalf of the user. For example, the user may be automatically notilied of products (or other items of interest) which may be identified to be of personal relevance at such time that a user approaches such items geographically. Physical proximity can also be used as an additional selection criteria lOr personalization (as a client filter or on the vendor's server) in conjunction with IS other modes of personalized user access to (and pre-fetching ol) information as is performed during querying (searching) or browsing modes (including personalized menu or "portal" navigation).

An article Matching User's Semantics with Data Semantics in Location-Based Services" by Shijun Yu et al. SME'05. May 9, 2005, Ayia Napa, Cyprus, shows combining location-based services with an event notification system. When a tourist is walking around in a city, and stops at a sight (e.g. a castle), the system provides him/her with general information (e.g. general information about the castle). It also considers the user profile to deliver niore specific information (e.g. architectural information about the castle). It looks fOr relevant information in databases of scheduled events and spatial databases and notifies the user (e.g. there is a concert tonight in the castle). It compares the situation with the relevant part of the user's history (e.g. you saw a similar castle in London last week).

US patent application 20060238381 shows a single repository for capturing.

connecting, sharing, and visualizing inibmiation based on a geographic location.

Community based content and location information provided by sources combined with features such as real-time social networking are joined to provide one-stop answers about what is going on at a particular place. Combining location, history, time, and comniunity information facilitates an automated selection of local events, restaurants, places and more based on a user's (and people like the user) historical set of decisions. According to another embodiment, temporal attribution allows users to get information on past, current, or future (forecast) data. Discrete time-based inlormation is supported. such as event as well as real-time or near real-time data (e.g.. traffic, weather, river, or tide levels, ote tallies, observe suninioonistars.

seasons, and other temporally tagged information that can be modeled, archived, or forecast). Also supported is historical data to allow applications to perform time filters of projections. such as historical and predictive traffic data.

A position is not saved each time an event is recorded but rather a location-tracking service maintaining a "breadcrumh trail" of the user location at a predefined fixed interval (e.g.. 5 minutes, 15 minutes, 30 minutes, . . . ). This information can later be utilized to determine the location of any time-stamped information, for example, a document edit timestamp can reveal where a specific change to a document was performed (and possibly prompt memory recollection of why an edit was made). The time-stamp information can also facilitate time-shilling functionality that allows a user to icw not only current data, but also historical information and predicated future information. For example, the user can time-shill to see hat the weather will be like tomorrow, based on historical forecasts or to time-shift to an earlier date to accurately observe how the weather behaved.

US application 20070061 245 describes tracking a mobile communication facility, presenting search results based at least in part on a first location. An implicit search niay be initiated based on location and/or time and the implicit search may be designed to provide locally stored information that may be later used and is relevant to the location andlor time. For example. a person may Ily to a new location (e.g. Kansas City) and upon recognition that the phone is in a new location, a search may be launched. The search may be adapted to request directories such as yellow pages, white pages. maps, classified advertisements, hotel information, rental car information, public transportation information, services locations, good provider locations, restaurants, or other directory information.

US patent application 2003 135582 shows a context-aware search service in which a user enters a query, the query is customized with contextual information including the real-time condition and situation of the user, and a result to the customized query is relayed to the user in a customized manner. Additional information is added to the query to indicate the context of the customer or terminal. This addition information can include context information regarding at least the customer, the environs (spatial and temporal) of the user or the speci tic temiinal from which the user is accessing the virtual operator, and the particular characteristics of the terminal. The information about the user may include. for example, the identity olthe user, the activity in hicli the user is engaging. or the locomotive means of the user (e.g. is the user traveling by loot or via automobile). The environs of the user or teniiinal may include, for example, the date/time of year. time of day, or nearby people or activities. The time may be. lhr instance, the time olthe request.

Summary of the Invention

An object is to provide improved apparatus or methods. In one aspect the invention provides: A system for recommending to a user, media content related to a visit away from a 1 5 hon-ic location, by the user, the system having: a context collector arranged to obtain at least some visit context information automatically without prompting the user for manual entry of the visit context information, the visit context information comprising at least a pattern of locations of the user during the visit, and a recommender arranged to match after the visit, the visit context infomrntion from the context collector, and media content available for consumption after the visit, and arranged to present recommendations to the user after the visit, of which of the media content to consume, according to results of the matching.

Compared to above mentioned existing recommenders based only on using past media consumption or predetermined user preferences, the use of visit context information can enable more appropriate recommendations for later consumption. Furthermore the recommendations can be wider in scope since they may encompass media content relevant to things that the user was unaware ol These things can enconipass locations that the user was unaware of, or unknown characteristics of a location that the user was aware oC Compared to above mentioned known location dependent mobile search system, or the above mentioned known mediascape system, which are suitable for media content for consumption during the visit, it can be advantageous to be able to carry out the matching aller the isit, rather than during the visit, since then matching can then be based on multiple locations during the visit, giving a pattern olmovements. not usE a present location. Thus the recommendations can be more balanced and relate better to the overall visit. Furthermore, presenting recommendations alter the isit rather than during the visit. may be prelrred for a number of reasons. For instance, the user niay have more time. may be more comfortable, and in an enironmenE and a time more suited to consuming the media, such as being at the home location.

Furthermore, compared to any of the above mentioned mobile search systems which require user input of search query terms before they can start to search, it can be much more conenient for a user ii the user does not need to enter isiE context information manually, or work out how to search manually or work out what search query terms to start with at least. These advantages still apply even lithe user sometimes desires enhanced recommendations by entering some information manually. Recommender systems generally become more and more valuable as the amount of media content available becomes greater, and as less time is available for manual searching of' media content services, such as by reading lengthy broadcast programme listings, or web search result lists for example.

The above aspect of the invention can be enibodied with any additional features to the above-mentioned laturcs, and some notable additional tatures are set out in dependent claims and some are described in more detail below.

Other aspects include corresponding niethods of providing a recommender service and methods of using such a service. The methods of use aspects can help to enable direct infringement or inducing of direct infringenient in cases where parts of the system are located outside the jurisdiction covered by the patent, as is feasible with many such systems. yet users are using the system and gaining the benefit, from within the jurisdiction. Other advantages than those set out will be apparent to those skilled in the art, particularly in comparison to other prior art. Any of the additional features can be combined together. and combined with any of the aspects, as would be apparent to those skilled in the art. The embodiments are examples only, the scope is not limited by these examples. and many other examples can be conceived within the scope of the claims.

Delinitions A visit is a period of time between a user leaving a home location and returning to ii, or arriving at a new home location, or a number of such periods of time.

A "home location" is a place where the user is permanently or temporarily located before or after visits, such as a user's home, a user's oflice, a hotel where the user is staying. and max' be defined by the user beforehand, or defined instantaneously by the user, for example when the user rests at a café, or is waiting for a train, they can define that location as a home location.

To match" is intended to mean finding any kind ol'relationship between visit context information and media content, such as to find matches in geographical names, nanies olpoints of interest. matches in terms of proximity of' location coordinates, matches in terms of rules defining relationships, to find similar previous visits, by finding statistical or other correlations of patterns with previous isits, by the user or other users, and then looking up what media content choices were made following such similar isits.

To "present recommendations" is intended to encompass presenting in any way, such as by presenting visually text, images or icons, or by audio, or any other way, and is intended to encompass recommendations in any order, or ordered or ranked by any criteria, such as a list ol'titles or summaries, or pages of' such lists, or any other way.

Brief Descrit,tion of'the Figures Specific embodiments of the invention will now be described, by way of' example.

with reference to the accompanying Figures, in which: Figure 1 shows a first embodiment according to the invention.

Figure 2 shows an embodiment in which the context collector and recommender are implemented as client side software at a home location, Figure 3 shows an embodiment in which the context collector and recommender arc implemented as server side software at a remote centralised location, Figure 4 shows an embodiment in which the context collector is implemented as client side software, and the recommender is implemented as server side software at a remote centralised location, Figure 5 shows a time chart for an embodiment showing actions of difThrent parts, Figure 6 shows an iniplementation of a context collector l'or use in any of' the above embodiments, Figure 7 shows an implementation of processing of video or images by the context collector, and Figure 8 shows an implementation of a recommender lbr use in any of the aboe embodiments.

Description of Embodiments

Introduction:

As will be described below, in at least some embodiments, parh of known recommender systems and known context-based mobile applications can be combined in novel ways. Some embodiments are notable in that the collection of contextual information and its use is automatic and transparent rather than explicit and directed by the user. Some embodiments of the recommender system can make and act upon media recommendations for home consumption based on a mobile user's context.

Some embodiments involve the following: IS -A mobile device that is able to monitor and act upon aspects of the user's context (for example their location), -A means of transIrring contextual inl'ormation to a recommender system in real time or at some later point, and -A recommender part that can identify, recommend and (possibly) operate on media content such as media objects, based on matching which can involve an analysis of the transIrred visit context information.

A simplilied example application can be summarised as follows: A user visits a popular historic site such as the Tower of London and walks around.

Her mobile device tracks her location. On returning home, she docks her mobile device which automatically uploads a record of her journey to her home recommender system. The system analyses this information and uses its knowledge base to identify video about the Tower of London and related topics, downloads that video from sonic service, and prepares it for presentation on the user's television. After taking off her coat and making a cup of tea, the user settles down in front of the TV to review her photos and watch a recommended program on Henry VIII.

A variation is that contextual information could be transferred to the recommender system in real time over some ireless network connection (for example Wi-Fi, or GSM) or a Bluetooth TM link for example. Another variation is that the recommender system could be provided by an online service rather than a home device.

Notable is that the collection of contextual data can be done automatically and without the user's conscious intervention. All ol' the user's previous context (for example here they have been over a period of time) can be available to drie a recommendation engine.

Some additional features, and comments on why they are notable Additional features of particular embodiments can include the recommendations being of media content available for consumption at the home location after the visit. The home location is typically more convenient for the user, since the user may have more time, it may be more comfortable, and typically has facilities or devices more capable than mobile devices in terms of screen size, audio quality, processing power, storage capacity, communications bandwidth and so on.

The context collector can be arranged to obtain at least some of' the context information from information retrieved from a mobile device carried by the user during the visit. This is notable for enabling the system to be more self-contained and not need to rely on external remote services.

Another such additional fCature is the context collector being arranged to obtain at least some ol' the context inforniation from information retrieved from one or more remote centralised providers. This is notable for enabling the system to be niore transparent to the user and not need to rely on the user's mobile device. Thus it flay make the system more universal by reducing or avoiding the need to customise the system to suit different mobile devices, and avoid or reduce the need to connect the mobile device to the context collector. Whether the context collector is located at the home location or at a remote location, it will usually be easier to connect it to remote providers than connect to the user's mobile device.

The context collector can be arranged to retrieve and use as a basis for obtaining the visit context inlormation, any one or more o! sounds, images or video recorded by the user during the visit, records of user's mobile network device activity stored by the device or stored by the mobile network, records of satellite navigation system activity, stored by the navigation system or stored by the user's navigation device, records of purchasing systems used by the user, records of' a vehicle tracking system.

These are some ways in which context information can be obtained automatically from other systems used during a visit without needing additional actii1y by the user to input visit context information manually to the system.

The context collector can be arranged to retrieve information having implicit location information, and use it to derie isit context inl'omiation having explicit location information. This can iiiake the matching easier or more accurate.

The explicit location information can comprise any one or more of location coordinates, points of interest, keywords of content type.

At least some of' the media content can hae tags indicating location, the recommender being arranged to derive location information from the context information and match the tags and the location inlbrmation. This can make the matching easier or more accurate. In some cases it niay be easier to match by location.

In other cases it may be easier to con'ert context information to a keyword or media content category. e.g. history or art, and use that for the matching rather than deriving IS a location from the content.

The media content can coniprise television channels having corresponding electronic programme guide information, and the recommender can be arranged to receive the electronic programme guide information and carry out the matching using the electronic programme guide information. This is currently a valuable application of recommendation engines.

The media content can comprise content accessible and downloadable over the Internet, and the recomniender can be arranged to determine the matches by automatically deriving keywords from the context information, sending the keywords as a search quety to a third party search engine, receiving search results from the third party search engine, and selecting at least some of these as matches. This is also an increasingly valuable source of media content such as web pages, sound and video clips and so on. but currently it is time consuming and needs much user input to find relevant material.

The context collector can comprise predominantly client side software arranged to run on a user's device. This may be easier and more cost effective to implement, as much of the information needed may be accessible from the users mobile device or from a home computer of the user, or both. Another advantage is that it can enable more privacy ii only the users devices have detailed information about the users visit, and the context collector can ensure that only more generalised or liltered or impersonal information is passed on.

The recommender can comprise predominantly server side software arranged to run on a remote server as a service for many users and arranged to send the recommendations to the corresponding users. This may he more convenient to enable the recommender have easier access to large databases and more processing power.

Also it can make the systeni more suitable for internet connected de ices and make it easier to upgrade and maintain centrally l'or many users.

The recommender can be arranged to access a stored user proIle and determine the recommendations based additionally on the profile. This can make the recommendations more appropriate.

The context collector can be arranged to identify points of interest along a path taken by the user, and determine relevant points of interest according to how close are the points of' interest to the path, and according to a duration the user spent close to that IS point ol' interest, and use the relevant points of interest in the matching. This can help make the matching easier in sonic cases, since keywords or media categories can often be derived from points of interest to augment matching based on location.

The system can be arranged to feed the recommendat ions to another system or service, for incorporation with other information to be presented to the user. This can make it easier fbr a user to see the recommendations more conveniently with less effort if' they are incorporated with other information that is consulted regularly such as email lists, or electronic calendar or electronic program listing for example.

The system can be arranged to prompt the user for manual entry of' further visit context inI'omiation for use as basis for enhanced l'urther recommendations. This can help combine some of' the advantages of user input to get more appropriate results, when needed, with the advantages of' the initial recommendations being made without user input.

Another aspect provides a corresponding method ol'providing a recommender service, l'or recommending to a user, media content related to a visit by the user away from a home location. the method having the steps of obtaining at least some visit context infomiation automatically without prompting the user l'or manual entry of the visit context information, the visit context information comprising at least a pattern of locations of the user during the visit, aficr the isit. matching the visit context information and media content available for consumption aller the visit, and presenting recommendations to the user afier the visit, of which of the media content to consume, according to results of the matching.

Another aspect of the invention provides a corresponding method of using a recommender service for recommending to a user, media content related to a visit by the user away from a home location, the method having the steps of enabling a mobile device to record isit context information comprising at least a pattern of locations oithe user during the visit, making the visit using the mobile device, returning to a home location, sending the recorded visit context information from the mobile device, to the recommender service, and receiving from the recommender service a presentation of recommendations based on IS the visit context information, olmedia content for consumption at the home location.

The enabling of the mobile device can be carried out earlier so that it continuously tracks, or it can be switched on or off just for a given visit or part of a visit. The enabling of the system to receive the visit context information can involve docking the mobile device at honie, or previously setting up the system and the mobile device so that the context information is transferred wirelessly from the mobile device during or aller the visit for example, using a cellular mobile network or a home ircless local area network for example.

Fig I, a first embodiment Figure 1 shows a schematic view of some of the principal elements olan embodiment of a system for recommending to a user, media content related to a visit away from a home location, by the user. In ligure 1, a context collector 220 is coupled to computer readable records 100 of a user's visit., the context collector is arranged to obtain at least sonic visit context information automatically without prompting the user for manual entry of the visit context inforniation, the visit context infomiation comprising at least a pattern of locations of the user during the visit. The visit content information is led to a recommender 230. This is arranged to match alter the visit, the visit context information ironi the context collector, and media content 120 available for consumption after the visit. The recommender is arranged to present recommendations to the user 140 after the visit, of hicli of the media content to consume, according to results of the matching. The recommendations can be presented at the lionie location or elsewhere, such as on a user's mobile device. The user can then make a selection [mm the recommendations, to consume the media content, for example by aiching a TV program, or film, at the home location or else here aller the visit. The media content can be downloaded over fixed links (such as cable, satellite or wired phone line to the home location, or conceivably donloaded to a user's mobile device, which typically has lower bandwidth than the fixed links.

Figs 2,3,4. further embodiments These figures show embodiments showing some elements implemented at a home location and other elements implemented away from the home location. In figure 2, the context collector 220 and recommender 230 are implemented as client side solware at a home location. Figure 3 differs in that it shows an embodiment in which the context collector and recommender are implemented as server side sofiware at a remote ceniralised location. Figure 4 dilIers in that it shows an embodiment in which the context collector is implemented as client side software, and the recommender is implemented as server side sofiware at a remote ceniralised location.

In fig 2 the context collector and the recommender are implemented as software running on a personal computer PC 225. Other architectures are conceivable. The context collector is coupled to receive information from the user's mobile device 210.

In this case, the mobile device is shown with a number of optional components such as a satellite navigation thcility (sat nay) which couldbe in the form of'a GPS or any other type. The mobile device also can have a camera 3 15, and other sensors 325. The mobile device can have a cellular mobile phone function ako. The other sensors could include for example bio sensors to detect users heart rate, skin temperature or other indicators of health or of emotional response to their surroundings during the visit.

These bio-factors can be part ol the visit context info and can help indicate which parts of a isit alThcted a user. The context collector can be used with mobile devices having any combination of these functions, or having other functions such as beacon sensors for detecting beacons which transmit location information, such as RF, IR, or other types of beacons. The mobile device could also have WiFi and Bluctooth connectivity, and detect its location from corresponding transmitters.

The context collector is also sho n coupled to remote providers 270 of implicit location information. These can be centralised database systems holding information about activities of many users. In some cases there ill be priacy and commercial conlidentiality considerations, but if these can be overcome, then the context collector could he used to retrieve information about the users own activities. The illustration shows credit card systems 2() which would have information about identities or locations of points of sale and times of transactions by the user at those locations.

Mobile networks 290 could have databases of calls and of roaming locations of the users mobile phone, including logs recording times of handovers between base stations and the identities of the base stations for example. Similarly, sat nay systems 300 could have logs of users paths, if there is sonic arrangement for downloading them from a user's device. The context collector could interrogate any or all of these 1 5 remote providers, or other types of remote providers.

The collected information can be processed by the context collector, for example to derive explicit location information from implicit information. The collector can optionally combine information from different sources, and make deductions from the combinations. The resulting visit context information is fed to the recommender 230.

The recommender niay access web based media content through an internet link to the world wide web WWW 320 and to 3rd party search engines 3 10. For a home based recommender, there may be a direct link within the house to digital TV systems to access Electronic program guide EPG information. Alternatively, or aswell, the EPG information can typically be obtained on-line, as shown in the figure by the EPG info being passed from the TV sources 340 to the WWW. The TV sources are shown as cable 350, satellite 360 and terrestrial 370, and any or all of these can be used.

Terrestrial TV can be analog or digital. Currently only digital terrestrial TV is transmitted with EPG, though conceivably the teletext data on the analog signals could be converted as used as EPG information. The user has a household digital TV system shown by the TV receiver,'decoder 260, with an EPG processor 250, coupled to a TV/display 240. The TV receiver/decoder could be part of a PVR for example.

The PC and TV can have separate displays, or be linked. For example the PC could be integrated with the PVR circuitry or the PVR can be integrated within the PC. Many dilicrent architectures can be conceived for how these elements are implemented or coupled together, to achiee the basic functions of internet access for the recommender, and access to EPG information Ibr the recommender, and a display for presenting the recommendations to the user.

In sonic embodiments, the recommendations can be presented on the same screen as is used by the user for consuming the media, such as the users home TV. or a user's desktop PC display. Of course the recommendations can he presented elsewhere such as on the users mobile device. The recommendations can be presented as hyperlinks to enable the listed recommendations to be selected and consumed immediately with one click by the user, or in any other way convenient to the user.

Fig 3 has similar Iatures to those of hg 2, shown with the same reference numerals.

but differs in having a remote recommending service 275 run as server side software on conventional server computing infrastructure. The service has a context collector and recommender which can be implemented in sohiware and can operate in a similar I 5 lishion to the corresponding elements shown in fig 2 at the home location. It may be harder for the service to access information from the user's mobile device, but this is still possible with an internet connection over a wireless network for example. The recommender is shown coupled to a web browser 380 in the home location. This web browser is one way of accessing the remotely generated recommendations. The web browser can run on a home location PC, or other conventional computing infrastructure such as may be integrated into a PVR or similar, or on the user's mobile device.

The user can use the web browser to access the recommender over the Internet. The web browser will receive the recommendations, or alternatively the recommendations can be sent to the user's mobile, as shown by a dotted line. Another alternative is to fl.cd the recommendations to the TV sources, again shown by a dotted line. These systems typically have high bandwidth broadcast paths and low bandwidth paths for teding small amounts of data to individual subscribers or users. These low bandwidth paths can be used to pass recommendations to individual subscribers TV screens or to EPG circuitry at the honie location, for example, to enable display of recommendations incorporated with other EPG information.

Fig 4 shows a further embodiment similar to those shown in Figs 2 and 3, and similar features have corresponding reference numerals. In this case, the recommender is located remotely, and the context collector is implemented as client side software running on the user's computing infrastructure. This may be a PC at the home location, or the user's mobile device for example, or both. As discussed above, this may he easier or more cost effective to implement as much inforniat ion about the isi1 may be already present on the mobile device, or the users home computer if linked to the mobile device, for example for synchronisation using a docking station or short range radio link such as Bluetooth. The context collector can access information i-elated to the visit from the user's mobile de ice, and or from the remote providers, as discussed ahoe in relation to ligs 2 and 3.

In Fig 4. the home PC, 225 is used to run the context collector and a web browser for displaying the recommendations. Again the recommendations could be integrated with EPG info. if it is convenient for a user to see the recommendations as soon as the user switches on a display of EPG information for example.

Figure 5, time chart summarising actions of principal parts Figure 5 shows a time chart with time flowing down the chart, for an embodiment of the invention, which niay correspond to any of the above described embodiments, or another embodiment. Actions of the user are shown in the left column, actions of the context collector in the central column, and actions of the recommender in the right hand column. After the visit, the user docks the mobile device for example, to synchronise the device with a local computer, or starts the process in any way. The context collector then obtains visit context information automatically without prompting the user for manual input. This is exemplified by the context collector downloading from the mobile device images or video taken during the visit, records of mobile calls or cellular roaming activity such as which base stations were contacted. More precise location information than merely knowing the mobile was in range of a given cell of the network may be obtainable by known triangulation methods. These involve obtaining indications of signal strength from a number of base stations, and the use of known triangulation techniques to establish position relative to these base stations. The context collector can also collect information on what web pages were browsed during the visit, and can access remote providers as discussed above, for further inlbmiation. The information can be processed to derive explicit location information from implicit information, for example by image processing the images to identify known locations, by looking up locations of cells based on cell ID, by retrieving web pages browsed, to look for location infbrmation, and so on. The location information can be collated, correlated to fill in a pattern of' locations and associated timings f'or example. associated points of interest, and so on.

Results can be output in any form suited to the recommender, for example a ranked list of locations, ranked by time spent or by user profile, or any other f'actors.

The recommender is arranged to match afler the visit, the visit context inf'ormation from the context collector with media content. This can involve for example using rules to deduce names or points of' interest, searching for web based media using location names or points of interest as query terms, searching EPG information for matching location tags or matching words in programme titles. Another matching approach is to use a database of prior recommendations or prior selections made by users after visiting given locations, and find prior which of them matches most closely the pattern of' locations of the present visit. Either way, the recommender can output a ranked list to the user. The user can then select f'rom the recommendations, or can be prompted to enter more visit information to get enhanced recommendations.

Figures 6 and 7. an implementation of'a context collector Figure 6 shows a flow chart of' steps for an example implementation of a context collector 220 for use in any of' the above embodiments, or in another embodiment of the invention. At step 200 a start time of' the visit is determined. This can be done by detecting when a user left the home location, when any location information away from the home location is detected, or by prompting the user, or in any other way. The time of start of the visit can be used to help limit the search for visit context information. A number of' dif'f'erent paths of' actions can follow as illustrated. It is optional which or how many of' these paths are used in a given implementation. At step 420, roaming records are retrieved from the mobile device or network. Times of' messages and identities of' base stations are extracted at step 430. At step 440, base station IDs are converted to location coordinates, such as latitude and longitude.

Records of a user's path are retrieved from a sat. na. system at step 410. This can be from the mobile device. which could be a hand held or an in car device for example, or possibly from a central database. Records from other location based systems can also be used, such as beacon transmitters, RF-ID tracking systems ehicle tracking systems such as vehicle number plate recognition systems, tracking systems based on image recognition or face recognition by CCTV cameras. At step 41 5, bio and other records are retrieved for the lime period of the visit. These can include for example heart rate and skin temperature or any other indicator of a user's interest, excitement.

exercise level and so on.

At step 450, purchasing records are retrieved. This can be from a credit card system.

or from prepayment cards and so on. Purchasing records on a mobile device or card may pros ide IDs of the points of sale in retail establishments such as shops, and so it may be necessary to look up locations based on these IDs, as shown at step 460. At step 490, image processing of video or images taken during the isit is carried out. An example implementation of this step is shown in figure 7.

All this activity can take place during the visit, or aller the visit, or both. At step 470 the context collector can use the inlbmmtion derived or retrieved in the various ways described, to look up points of interest and geographical names for example located IS within a predetermined range of the path of the user, depending on various criteria such as the time spent by the user at the different locations, such as a user profile, or any other criteria. At step 480. keywords of content type related to locations or related to points of interest for example can be derived. Furthermore, collating of the information retrieved or derived from the different sources can be carried out, and any inconsistencies identilied and corrected for example. Timings and other information can be deduced, for example using rules such as if heart rate is in range x, and rate of change of location is within range y, assume user is walking". A ranking of the ditierent locations along the path can be derived, based on how much time was spent, and based on other factors. The visit context infOrmation can then be sent to the recommender and can be in the form of a pattern of' locations, such as a list of coordinates, optionally with associated information such as any one or more of ranking, timings, and keywords such as points of interest and content categories associated with sonic or each of the locations.

Figure 7 shows an example of how the video,' images can be processed, for use in the implementation of figure 6 or for other uses. At step 500, video and/or images are retrieved from the mobile device. The inlages such as frames of the video are searched for those showing landscapes such as views of buildings or signs or any other geographical content. At step 520 the image processing involves searching for text in signposts, nameplates etc. to identil' points of interest or geographical names. At step 530, the images such as the frames are sent to an inlage search engine to search for matches with images in image libraries, again to identit' location, points ol'interest or geographical names. At step 540. location coordinates are derived for any of the identilied points of interest or geographical names. The coordinates and the other information can be fed to the rest of the context collector for further processing such as collating with other visit information a described above.

Figure 8 implementation o ía recommender An implementation of a recommender 230 is shown in ligure 8 for use in any of the embodiments described, or in other embodiments. The recommender has a recommendation engine 580, a database 590 of rules for matching, a database 570 of statistics of other users choices aller prior visits. This can be indexed by locations or patterns of locations of the prior visits. The recommendation engine is coupled to IS receive inputs of visit context information lroni the context collector, and optionally from manual user input for example. This visit context info can comprise for example location coordinates, time spent at the coordinates, ranking of the locations, associated other information such as location names, geographical names, point of interest, other visit information such as bio data and environmental information such as light level or noise level, or weather for example. The recommender is also coupled to receive web based media content, in response to queries that it generates and sends to third party search engines for example. The recommender is also arranged to receive EPG information, and may have a part 560 for deriving location information suc has coordinates, points of interest and geographical names for example, from the EPG information. A led from a user profile database 590 is also shown. The recommendation engine can produce a ranked list of recommendations based on some or all the inputs. The engine may operate according a statistical approach based entirely or partly on previous user's choices, without reasoning why the recommendations relate to the locations. Alternatively or as well, the engine may operate according to a reasoning approach using rules for matching, which may be based on reasoning such as interest in the tower of London could suggest an interest in castles, and in history of Kings and Queens, so search for programmes or films with any of these words in the titles or summaries" . The rules can be developed manually,

-I

or sonic can be deduced li-om the statistical database 570. The database can be populated by feedback of prior selections made by other users, together with the patterns of locations of their visits, and their user profile information for example. The recommendation engine and design of the databases can be carried out l'olloing established practice For recommendation engines for other applications such as recommending purchases based on prior purchasing patterns by similar users.

The solhare can be implemented using any conventional programming language, including languages such as C, and compiled following established practice. or object oriented languages. The servers and network elements can be implemented using conventional hardware with conventional processors. The processing elements need not be identical, but should be able to communicate with each other. e.g. by exchange oFIP messages.

Other variations can be conceived within the scope of the claims.

Claims

I. A system lbr recommending to a user, media content relaled to a visit away from a home location, by the user, the system having: a context collector arranged to obtain at least some visit context information automatically without prompting the user for manual entry ol' the visit context information, the isit context information comprising at least a pattern ol' locations of the user during the visit, and a recommender arranged to match aller the visit, the isit context information from the context collector, and media content available for consumption alter the visit, and arranged to present recommendations to the user alter the visit, of which ol'the media content to consume, according to results of the matching.
2. The system of claim I, the recommendations being of media content available for consumption at the home location alter the visit.

IS
3. The system of claim I, the context collector being arranged to obtain at least some of the context infomiation from information retrieved from a mobile device carried by the user during the visit.
4. The system of' claim I, the context collector being arranged to obtain at least some of the context information from information retrieved from one or more remote centralised providers.
5. The system of claim I. the context collector being arranged to retrieve and use as a basis for obtaining the visit context information, any one or more of sounds, images or video recorded by the user during the isit. records of user's mobile network device activity stored by the device or stored by the mobile network, records of satellite navigation system activity, stored by the navigation system or stored by the user's navigation device, records of purchasing systems used by the user, records of a vehicle tracking system.
6. The system oiclaini 5. the Context collector being arranged to retrieve information having implicit location information, and use it to derive visit context iniomiation having explicit location information.
7. The system of claim 6. the explicit location information comprising any one or more oI location coordinates, points of interest, keywords of content type.
8. The system of claim I, at least sonic of the media content ha ing tags indicating location, the recommender being arranged to derivc location information from the JO context information and match the tags and the location information.
9. The system of claim 1, the media content comprising television channels having corresponding electronic programme guide information, and the recommender being arranged to receive the electronic programme guide information and carry out the matching using the electronic programme guide information.
10. The system of claim I. the media content comprising content accessible and downloadable over the Internet, and the recommender being arranged to determine the matches by automatically deriving keywords from the context information, sending the keywords as a search query to a third party search engine, receiving search results from the third party search engine, and selecting at least sonic of these as matches.
11. The system of claim I. at least the context collector comprising predominantly client side software arranged to run on a user's device.
12. The system of claim 1, at least the recommender coniprising predominantly server side software arranged to run on a remote server as a service for many users and arranged to send the recommendations to the con-esponding users.
13. The system of claim I the recommender being arranged to access a stored user profile and determine the recommendations based additionally on the profile.

_,, , -
14. The system of claim 1. the context collector being arranged to identify points of interest along a path taken by the user, and determine relcant points of interest according to ho close are the points of interest to the path, and according to a duration the user spent close to that point of interest, and use the relevant points of interest in the matching.
15. The system ofelaim I, arranged to feed the recommendations to another system or service, for incorporation with other information to be presented to the user.

It)
16. The system ofclaini 1 arranged to prompt the user for manual entry of further visit context information lbr use as bask fhr enhanced further recommendations.

1 7. A method of providing a recommender service, for recommending to a user, media content related to a visit by the user away from a home location, the method having the steps of: obtaining at least some visit context information automatically without prompting the user for manual entry of the visit context information, the visit context information comprising at least a pattern of locations of the user during the visit, after the visit, matching the obtained visit context inlbrmation and media content available for consumption after the visit, and presenting recommendations to the user afler the visit, of which of the media content to consume, according to results of the matching.

I. A method of using a recommender serb ice for recommending to a user, media content related to a visit by the user away from a home location, the method having the steps of enabling a mobile device to record visit context information comprising at least a pattern of locations of the user during the visit, making the visit using the mobile device, returning to a home location, sending the recorded visit context infomiation from the mobile device, to the recommender service, and receiving from the recommender service a presentation oirecomrnenthiions based on the isii Context information. of media content for consumption at the home localion.