AU2004237897A1 - Configuring a recommender of data items - Google Patents

Description

S&F Ref: 690158

AUSTRALIA

PATENTS ACT 1990 COMPLETE SPECIFICATION FOR A STANDARD PATENT Name and Address of Applicant: Actual Inventor(s): Address for Service: Invention Title: Canon Kabushiki Kaisha, of 30-2, Shimomaruko 3-chome, Ohta-ku, Tokyo, 146, Japan Julian Lewis Kerr Spruson Ferguson St Martins Tower Level 31 Market Street Sydney NSW 2000 (CCN 3710000177) Configuring a recommender of data items The following statement is a full description of this invention, including the best method of performing it known to me/us:- 5845c O CONFIGURING A RECOMMENDER OF DATA ITEMS 0 Field of the Invention The present invention relates generally to the fields of multi-media, information technology and computing and, in particular, to a method and apparatus for configuring a recommender of data items. The present invention also relates to a computer program 00 product including a computer readable medium having recorded thereon a computer program for configuring a recommender data items.

Background In recent times, particularly in the areas of multi-media, information technology and computing, people are being exposed to huge amounts of data every day. It is important that a particular person or user of such data is able to navigate through available data items efficiently, in order to determine a subset of data items that is more relevant to the particular user.

Conventionally, there have been two main methods of determining a subset of relevant data items from available data items. These two methods may be broadly referred to as "searching" and "browsing".

However, conventional searching methods require a considerable amount of data, such as keywords, search categories and other search conditions, to be input by a user in order to be effective. Further, conventional browsing methods only work well when the number of data items in a collection of data items being browsed is less than a certain size beyond which browsing efficiency starts to decline. Still further, most conventional browsing methods, do not take into account differences in individual preferences between different users.

Some conventional searching and browsing methods use "recommendation systems" (or 'recommenders') to generate a short list of data items that may be recommended to users. For example, a recommender for a television system may be used 690158.doc -2- O to rate television programs for recommendation to a user. The recommender may then o produce a list of television programs for recommendation to the user based on these ratings. The recommended television programs may be sorted in order of likelihood that the user would be interested in the television program sorted from the highest rated 5 television program to the lowest rated television program).

00 Recommenders may be configured explicitly by a user. Alternatively, a recommender may be configured automatically using user preference information Ocollected for the user. Such user preference information may be collected either explicitly by asking the user directly) or implicitly by interpreting the behaviour of the user) or both. Implicit user preference information may take the form of data items selected by a user for consumption. Such user preference information may comprise context information data items that were available for consumption but were not consumed). As an example, in response to a user watching an action film a recommender may be automatically configured to rate action films highly. Therefore, the recommender will recommend action films to a user. However, if a romance film was broadcast at the same time as the user watched the action film, the recommender may be configured to rate romance films lower. The recommender may therefore be configured to recommend action films in preference to romance films.

Explicit user preference information may take the form of a rating assigned to an item by a user. For example, a user may rate a television program highly where the television program is an action film starring Bruce Willis. In this instance, the recommender may be configured to recommend action films and/or films including Bruce Willis in the cast.

A recommender may also be configured using both implicit and explicit user preference information. The implicit and explicit user preference information may be used by the recommender to rate data items for recommendation to the user.

690158.doc -3- O Automated methods of rating data items, for recommendation by a recommender, o allow the recommender to be configured without requiring any effort to be made by a user. However, a disadvantage of these automated methods is that they may result in the recommender recommending data items that the user does not wish to consume, in a context where data items exist that the user does wish to consume. This may occur when 00 incorrect assumptions are made by the recommender when rating data items. For (Ni example, an automated method of rating data items may assume that items similar to an Sitem previously consumed by the user are suitable for recommendation. In this instance, the user who previously viewed a Bruce Willis film might be interested in being recommended other films where Bruce Willis is a cast member. The user might also regularly watch news. However, the user may not be interested in being recommended other news television programs and therefore, news television programs should not be rated highly.

In order to overcome the disadvantages of the above methods, one conventional method of configuring a recommender employs a user to assess the configuration of the recommender. In this instance, if the configuration of the recommender is determined to be suboptimal, the user may modify the configuration to improve configuration of the recommender. For example, the configuration of the recommender may rate new television programs highly resulting in news television programs being recommended.

However, the user may not be interested in receiving this kind of recommendation.

Therefore, the user may modify the configuration such that news television programs are rated lower and will not be recommended in the future.

However, one disadvantage of requiring the user to modify the configuration of a recommender is that the configuration must be in a form that is interpretable by the user.

For example, the user will not be able to interpret the configuration of a recommender that uses a trained neural network to generate recommendations.

690158.doc -4- O Thus, a need clearly exists for a more efficient method of configuring a o recommender of data items.

Summary It is an object of the present invention to substantially overcome, or at least ameliorate, one or more disadvantages of existing arrangements.

0 According to one aspect of the present invention there is provided a method of generating one or more scored concepts based on one or more data items, each of the data Oitems comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said method comprising the steps of: selecting one or more of the attributes associated with the data items; generating one or more concepts from the selected attributes, each of said concepts comprising at least one of the selected attributes; matching each of the generated concepts to one or more of the data items; statistically analysing the item scores of the data items matched to at least one of the generated concepts to generate a concept score for the at least one concept, wherein the generated concept score and associated concept represents a scored concept.

According to another aspect of the present invention there is provided a method of configuring a data item scorer, said data item scorer being adapted to generate an item score for each of one or more data items, the item score for an associated data item representing a predicted level of interest in the associated data item, said method comprising the steps of: generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; 690158.doc O statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one generated concept, the generated concept score and associated concept representing a scored concept; selecting one of said data items based on said scored concept; 0displaying one or more attributes associated with the selected data item; Ninputting a rating for the selected data item; and Oconfiguring the data item scorer using the inputted rating.

According to still another aspect of the present invention there is provided a user interface for configuring a data item scorer, said data item scorer being adapted to generate an item score for each of one or more data items, the item score for an associated data item representing a predicted level of interest in the associated data item, said user interface comprising: selection means for selecting one of said data items based on a scored concept; display means for displaying one or more attributes associated with the selected data item; and input means for inputting a rating for the selected data item, wherein one or more concepts are generated from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes, and the item scores of one or more of said data items that match at least one of the generated concepts are statistically analysed to generate said scored concept.

According to still another aspect of the present invention there is provided a method of determining a rated concept based on one or more data items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said method comprising the steps of: 690158.doc -6- O generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one 0 generated concept, the generated concept score and associated concept representing a N scored concept; Odisplaying a natural language description of the scored concept; inputting a rating for the scored concept based on the displayed description; and associating the inputted rating with the scored concept to determine the rated concept.

According to still another aspect of the present invention there is provided a method of configuring a data item recommender using a rated concept based on a plurality of data items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said method comprising the steps of: generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one generated concept, the generated concept score and associated concept representing a scored concept; displaying a natural language description of the scored concept; 690158.doc -7- O inputting a rating for the scored concept based on the displayed description; o associating the inputted rating with the scored concept to determine the rated concept; and configuring the recommender using the rated concept.

According to still another aspect of the present invention there is provided a method 0 of rating one or more data items for use in configuring a recommender of data items, each of the data items comprising one or more associated attributes values, said method comprising the steps of: selecting a concept from a set of concepts, each of said concepts comprising one or more of said data attribute values; selecting a data item based on the data attribute values of said concept; displaying one or more data attribute values associated with the selected data item; inputting a rating for the selected data item based on the displayed data attribute values for the data item; and storing the inputted rating and the selected data item for use in configuring the recommender.

According to still another aspect of the present invention there is provided a user interface for rating one or more data items for use in configuring a recommender of data items, said user interface comprising: first attribute display means for displaying one or more data attributes associated with a concept; data item display means for displaying one or more data items associated with the concept, the displayed data items being selectable and one or more of the displayed data items having been previously selected; second attribute display means for displaying one or more data attributes associated with at least one selected data item; and 690158.doc -8rating display means for displaying a rating associated with one or more of the data oattributes displayed in the second data attribute display means, wherein the displayed rating is input using the user interface.

According to still another aspect of the present invention there is provided an apparatus for rating one or more data items for use in configuring a recommender of data 00 items, each of the data items comprising one or more associated attributes values, said (Ni N apparatus comprising: Oconcept selection means for selecting a concept from a set of concepts, each of said concepts comprising one or more of said data attribute values; data item selection means for selecting a data item based on the data attribute values of said concept; display means for displaying one or more data attribute values associated with the selected data item; input means for inputting a rating for the selected data item based on the displayed data attribute values for the data item; and storage means for storing the inputted rating and the selected data item for use in configuring the recommender.

According to still another aspect of the present invention there is provided a computer program for rating one or more data items for use in configuring a recommender of data items, each of the data items comprising one or more associated attributes values, said program comprising: code for selecting a concept from a set of concepts, each of said concepts comprising one or more of said data attribute values; code for selecting a data item based on the data attribute values of said concept; code for displaying one or more data attribute values associated with the selected data item; 690158.doc O code for inputting a rating for the selected data item based on the displayed data attribute values for the data item; and code for storing the inputted rating and the selected data item for use in configuring the recommender.

According to still another aspect of the present invention there is provided an 0 apparatus for generating one or more scored concepts based on one or more data items, N each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said apparatus comprising: selection means for selecting one or more of the attributes associated with the data items; concept generation means for generating one or more concepts from the selected attributes, each of said concepts comprising at least one of the selected attributes; matching means for matching each of the generated concepts to one or more of the data items; analysing means for statistically analysing the item scores of the data items matched to at least one of the generated concepts to generate a concept score for the at least one concept, wherein the generated concept score and associated concept represents a scored concept.

According to still another aspect of the present invention there is provided a computer program for generating one or more scored concepts based on one or more data items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said program comprising: code for selecting one or more of the attributes associated with the data items; 690158.doc O code for generating one or more concepts from the selected attributes, each of said o concepts comprising at least one of the selected attributes; code for matching each of the generated concepts to one or more of the data items; code for statistically analysing the item scores of the data items matched to at least one of the generated concepts to generate a concept score for the at least one concept, 0 wherein the generated concept score and associated concept represents a scored concept.

(Ni According to still another aspect of the present invention there is provided an Capparatus for configuring a data item scorer, said data item scorer being adapted to generate an item score for each of one or more data items, the item score for an associated data item representing a predicted level of interest in the associated data item, said apparatus comprising: concept generation means for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; analysing means for statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one generated concept, the generated concept score and associated concept representing a scored concept; selection means for selecting one of said data items based on said scored concept; display means for displaying one or more attributes associated with the selected data item; input means for inputting a rating for the selected data item; and configuring means for configuring the data item scorer using the inputted rating.

According to still another aspect of the present invention there is provided a computer program for configuring a data item scorer, said data item scorer being adapted to generate an item score for each of one or more data items, the item score for an 690158.doc -11- O associated data item representing a predicted level of interest in the associated data item, said program comprising: code for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; 0 code for statistically analysing the item scores of one or more of said data items that Nmatch at least one of the generated concepts to generate a concept score for the at least Sone generated concept, the generated concept score and-associated concept representing a scored concept; code for selecting one of said data items based on said scored concept; code for displaying one or more attributes associated with the selected data item; code for inputting a rating for the selected data item based on the displayed data attribute values; and code for configuring the data item scorer using the inputted rating.

According to still another aspect of the present invention there is provided an apparatus for determining a rated concept based on one or more data items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item. said apparatus comprising: concept generation means for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; analysis means for statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one generated concept, the generated concept score and associated concept representing a scored concept; 690158.doc -12- O display means for displaying a natural language description of the scored concept; o input means for inputting a rating for the scored concept based on the displayed description; and association means for associating the inputted rating with the scored concept to determine the rated concept.

0According to still another aspect of the present invention there is provided a N computer program for determining a rated concept based on one or more data items, each Oof the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said program comprising: code for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; code for statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one generated concept, the generated concept score and associated concept representing a scored concept; code for displaying a natural language description of the scored concept; code for inputting a rating for the scored concept based on the displayed description; and code for associating the inputted rating with the scored concept to determine the rated concept.

According to still another aspect of the present invention there is provided an apparatus for configuring a recommender using a rated concept based on a plurality of data items, each of the data items comprising one or more associated attributes and an 690158.doc -13 O associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said apparatus comprising: concept generation means for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; 0analysis means for statistically analysing the item scores of one or more of said data (Ni items that match at least one of the generated concepts to generate a concept score for the Oat least one generated concept, the generated concept score and associated concept representing a scored concept; display means for displaying a natural language description of the scored concept; input means for inputting a rating for the scored concept based on the displayed description;associating the inputted rating with the scored concept to determine the rated concept; and configuration means for configuring the recommender using the rated concept.

According to still another aspect of the present invention there is provided a computer program for configuring a recommender using a rated concept based on a plurality of data items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said program comprising: code for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; code for statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least 690158.doc -14- O one generated concept, the generated concept score and associated concept representing a o scored concept; code for displaying a natural language description of the scored concept; code for inputting a rating for the scored concept based on the displayed description; 00 code for associating the inputted rating with the scored concept to determine the Srated concept; and code for configuring the recommender using the rated concept.

Other aspects of the invention are also disclosed.

Brief Description of the Drawings Some aspects of the prior art and one or more embodiments of the present invention will now be described with reference to the drawings and appendices, in which: Fig. 1 shows an example of a software architecture for implementing the described methods; Fig. 2 shows a program description; Fig. 3 shows a rated program; Fig. 4 shows a concept; Fig. 5 shows a scored concept; Fig. 6 shows the updateRecommender process of Fig. 1; Fig. 7 shows the getScoredConcepts process of Fig. 6; Fig. 8 shows the seekViewerFeedback process of Fig. 6; Fig. 9 shows a feedback dialogue; Fig. 10 shows a rated concept; Fig. 11 shows another example of a software architecture for implementing the described methods; Fig. 12 shows the seekViewerFeedback process of Fig. 11; 690158.doc Fig. 13 shows another feedback dialogue; o Fig. 14 shows a recommender search and configuration tool; Fig. 15 shows a recommender search and configuration tool user interface; e¢3 Fig. 16 shows the removal of a selected concept from the user interface of Fig. Fig. 17 shows a television system upon which the methods described herein may be 00 r-practiced; (Fig. 18 is a schematic block diagram showing the internal configuration of the settop-box of Fig. 17; and Fig. 19 shows a process for determining a score for the program description.

Detailed Description including Best Mode Where reference is made in any one or more of the accompanying drawings to steps and/or features, which have the same reference numerals, those steps and/or features have for the purposes of this description the same function(s) or operation(s), unless the contrary intention appears.

It is to be noted that the discussions contained in the "Background" section and that above relating to prior art arrangements should not be interpreted as a representation by the present inventor(s) or patent applicant that such documents or devices in any way form part of the common general knowledge in the art.

Methods described herein may be used to convert an arbitrary recommender configuration into a user-interpretable representation of the configuration. Such a userinterpretable representation may be conveyed to a user so that the user may evaluate the configuration of the recommender. The configuration may then be improved via feedback from the user.

The principles of the methods described herein have general applicability to modeling user preference information and recommending data items to a user. However, for ease of explanation, the steps of the preferred method are described with reference to 690158.doc 16- O recommending television programs to a single user or viewer of such television programs.

o It is not intended that the present invention be limited to the described methods. For example, the invention may have application to recommending many other products to C€3 the user. Further, a single user may also represent a plurality of users. For example, the single user may represent an organisation or even an animal.

00 Fig. 17 shows a television system 1700 upon which the described methods may Sbe practiced. The system 1700 comprises a set top-box 1701 coupled to a television set S1712 a digital television set). The system 1700 also comprises a remote control unit 1705, which may be programmed for selecting a television program for display on the television set 1712. As seen in Fig. 17, the remote control unit 1705 may comprise directional navigation buttons 1706 and function buttons 1707. The buttons 1706 and 1707 of the remote control unit 1705 may take any form. For example, one or more of the buttons 1706 and 1707 may be in the form of a defined area printed or formed on a surface of the remote control unit 1705 and which may be selectable by a user in a similar manner to the buttons 1706 and 1707. The remote control unit 1705 may also take the form of a smart card (not shown) and smart card reader (not shown), where the smart card has a number of user selectable indicia (or areas) formed on a surface thereof. In this instance, the user selectable indicia may be arranged on the surface of the smartcard in a similar manner to the buttons 1706 and 1707.

The set-top-box 1701 may be used to interpret signals 1709 received from the remote control unit 1705 according to a press of one or more of the buttons 1706 and 1707 of the remote control unit 1705, permitting control events to occur within the system 1700. These control events may result in changes to the state of the system 1700 and/or appropriate reproduction on the television set 1712, as will be described below.

In the system 1700, the remote control unit 1705 may use a radio frequency or IR transceiver (not shown) to transmit the signals 1709 to the set-top-box 1701.

690158.doc -17- O Alternatively, the remote control unit 1705 may be hard wired to the set-top-box 1701, Svia a communications cable (not shown). Similarly, the set-top-box 1701 is shown in Fig.

17 coupled to the television screen 1712, via a communications cable 1704.

¢€3 Alternatively, instead of being hardwired, a further radio frequency or IR transceiver 1808 (see Fig. 18) may be used for communication between the set-top-box 1701 and the 00 television set 1712.

SFig. 18 is a schematic block diagram showing the internal configuration of the settop-box 1701. The set-top-box 1701 is essentially a scaled version of a conventional (,i computer system. Such computer systems may include IBM-PC's and compatibles, Sun Sparcstations or alike computer systems evolved therefrom. Alternatively, the described methods may be practiced using a conventional computer system such as those described above, connected to the television set 1712, for example.

The set top box 1701 typically comprises at least one processor 1805, a memory unit 1806, for example, formed from semiconductor random access memory (RAM) and read only memory (ROM). The set top box 1701 may also comprise input/output (0I/O) interfaces including at least an I/O interface 1813 for transmitting data to and from a television set, for example. The I/O interface 1813 may also be used to transmit data to and from another device such as a portable floppy disk drive, CD-ROM drive or even to and from a communications network the Internet). The input/output (IO) interfaces of the set top box 1701 typically also include an 1/O interface 1815 for an IR transceiver 1808. The IR transceiver 1808 may be configured for receiving and transmitting signals to and from the remote control unit 1705, for example. The components 1805, 1806, 1808, 1813 and 1815 of the set top box 1701 typically communicate via an interconnected bus 1804 and in a manner which results in a conventional mode of operation.

Intermediate storage of any data received from the remote control unit 1705 may be accomplished using the semiconductor memory 1806. Alternatively, the components 690158.doc -18- 0 1805, 1806, 1808, 1813 and 1815 of the set top box 1701 may be configured within the television set 1712.

Software programs implementing the methods described herein may be resident in C€3 memory 1806 and be read and controlled in their execution by the CPU 1805 of the set top box 1701. Intermediate storage of the software programs may be accomplished using 00 the semiconductor memory 1806, possibly in concert with the CPU 1805. In some instances, the software programs may be supplied encoded on a CD-ROM or floppy disk and downloaded to memory 1806 via the I/O interface 1813. Still further, the software programs may also be loaded into memory 1806 from other computer readable medium including magnetic tape, ROM or integrated circuits, a magneto-optical disk, a radio or infra-red transmission channel between the set-top-box 1701 and another device, a computer readable card such as a smart card, a computer PCMCIA card, and the Internet and Intranets including via email transmissions of information recorded on Websites and the like. The foregoing is merely exemplary of relevant computer readable media. Other computer readable media may able to be practised without departing from the scope of the invention defined by the appended claims.

The set-top-box 1701 may alternatively be implemented in dedicated hardware such as one or more integrated circuits performing the functions or sub functions of the described methods. Such dedicated hardware may include graphic processors, digital signal processors, or one or more microprocessors and associated memories.

The software implementing the methods described herein may be configured to learn the preferences of a user who views television programs broadcast using the system 1700, for example. As described above, the software may be resident in the memory 1806 and be controlled in its execution by the processor 1805. Fig. 1 shows an example of a software architecture 111, which may be used for implementing the described methods. The software architecture 111 comprises a recommender 101, a program scorer 690158.doc -19- O 105 and an updateRecommender process 108. The recommender 101, program scorer S105 and updateRecommender process 108 may receive input and/or output data to an Electronic Program Guide (EPG) 100, a viewing history 104, a salient attributes list 109 and a rating history 106, configured within memory 1806 and/or the hard disk drive 1810.

The recommender 101 provides a program description 103 describing a particular 00television program to the program scorer 105. The program scorer 105 returns a score S102 for the television program described by the program description 103 to the recommender 101. The score 102 indicates an expected level of interest of a user for that television program. The program scorer 105 is configured using the viewing history 104 and a rating history 106. The viewing history 104 and the rating history 106 may be configured within memory 1806, for example. The program scorer 106 may be implemented using a back-propagation method on a feed forward neural network, for example. The score 102 may be a real number greater than or equal to one and less than or equal to five for example. The viewing history 104 contains a history of television programs that have been watched by the user. The viewing history 104 may also contain a history of television programs that were not watched by the user. The rating history 106 contains a history of television programs that have been rated by the user.

The updateRecommender process 108 rates one or more television programs by initiating a dialogue with the user of the system 1700 and receiving feedback from the user. The feedback may be in the form of a rated program 107, as seen in Fig. 1. The feedback may be provided by the user using the remote control unit 1705. The updateRecommender process 108 writes the rated programs 107) to the rating history 106 configured within memory 1806 and/or the hard disk drive 1810. An example of the rated program 107 will be described below with reference to Fig. 3.

690158.doc O The updateRecommender process 108 also receives program descriptions 103) Sfrom the electronic program guide (EPG) 100. The EPG 100 may be configured within memory 1806 and contains program descriptions 103). Fig. 2 shows an example of ¢€3 the program description 103 and will be described in more detail below.

The program scorer 105 provides the updateRecommender process 108 with scores 00 102) for each of the program descriptions 103 contained in the EPG 100. The SupdateRecommender process 108 may also receive television programs from the rating history 106. The history of television programs that have been watched by the user may (,i also be accessed by the updateRecommender process 108. The updateRecommender process 108 will be described in more detail below with reference to Fig. 6.

The salient attributes list 109 may comprise a subset of data attributes present in the EPG 100 that are considered salient to representing user preferences. The salient attributes list 109 comprises data attributes used by the program scorer 105. These data attributes may comprise one of more of the following attributes: program channel; (ii) program genre; (iii) program rating classification; (iv) program content warnings; and program duration.

The program descriptions such as the program description 103 are contained within the EPG 100. As seen in Fig. 2, the program description 103 comprises attribute-value pairs 201). Each attribute value pair 201) comprises an attribute start), which references a particular property of a particular television program, and a value 20:30 May which conveys the value of that property for the particular television program. Therefore, in the example of Fig. 2, the television program described by the program description 103 starts at '20:30 on 5 May'. The other attributes of the program 690158.doc -21- O description 103 may comprise channel, title, genre, rating, warnings and duration. With Sthe exception of start time and channel attributes, the program description 103 may contain multiple attribute-value pairs for a single attribute.

C€3 As seen in Fig. 3, the rated program 107 comprises a program description 103) and a viewer rating 302. The viewer rating 302 comprises an integer value assigned 00 by the user of the system 1700. The viewer rating 302 indicates the level of interest of the N user in the television program described by the program description 103. The integer value may be any integer from the set of integers one to five where a higher value represents a higher level of user interest.

Fig. 4 shows an example of a concept 400. A concept 400 comprises a set of attribute-value pairs 401). The program description 103 is said to "belong" to a concept 400 if the attribute-value pairs of the concept 400 form a subset of the attributevalue pairs of the program description 103. For example, the program description 103 of Fig. 2 belongs to the concept 400, since the attribute value pairs 401) of the concept 400 form a subset of the attribute-value pairs of the program description 103.

Conversely, if the program description 103 belongs to the concept 400, the concept 400 is said to "cover" the program description 103. For example, the concept 400 covers the program description 103 of Fig. 2.

The number of attribute-value pairs that a concept contains may be referred to as the "description length", t, of the concept. For example, the description length of the concept 400 is three t In another example, the description length, t, of a concept may be equal to the sum of salience scores of the attributes 401) of the concept, where the salience score of an attribute is a real number indicating the level of salience of the attribute. For example, the genre attribute may have a salience score of three the rating attribute a salience score of two and the duration attribute a salience score one In this instance, the concept 400 of Fig. 5 has a description length of six t 6).

690158.doc -22- The concept 400 is a conjunction of attribute-value pairs 401). Alternatively, the concept 400 may comprise a conjunction and/or disjunction of attribute-value pairs.

d Fig. 5 shows a scored concept 500. The scored concept 500 comprises the concept 400 and a concept score 502. The concept score 502 may be equal to the mean of the scores 102) of program descriptions 103) in the EPG 100 that belong to the 00 concept 400. The scores 102) of program descriptions 103) in the EPG 100 Sare determined by the program scorer 105 of Fig. 1. Alternatively, the concept score 502 may be equal to the mode or median of the scores 102) of the program descriptions 103) of the EPG 100 that belong to the concept 400 of the scored concept 500. A television program "belongs" to the scored concept 500 if the television program belongs to the concept 400 of the scored concept 500. The scored concept 500 "covers" a television program if the concept 400 of the scored concept 500 covers the television program.

The updateRecommender process 108 will now be described in more detail below with reference to Fig. 6. The updateRecommender process 108 may be implemented as software resident in the memory 1806 of the set-top-box 1701 and being controlled in its execution by the processor 1805.

The process 108 begins by generating a scored concept power set 601 using a getScoredConcepts process 600. The scored concept power set 601 comprises one or more scored concepts 500). The getScoredConcepts process 600 will be described in more detail below with reference to Fig. 7. The scored concept power set 601 is passed to a getDialogConcept process 602, which selects a scored concept 603 from within the scored concept power set 601. The scored concept 603 is of the same configuration as the scored concept 500 and includes a concept 400) and a concept score 502).

The getDialogConcept process 602 selects the scored concept 603 having the highest score 502) among a subset of scored concepts in the scored concept power set 601.

690158.doc -23 O The subset of scored concepts comprises scored concepts 500) whose concepts o 400) do not exist in a dialog history 604 configured within memory 1806 and/or the hard disk drive 1810. The dialogue history 604 comprises a list of concepts 400) that have been used previously by a seekViewerFeedback process 605 as the basis of a 5 dialogue with the user of the system 1700. The scored concept 603 is passed to the 00 0 0 seekViewerFeedBack process 605, which initiates a dialogue with the user. The C1 seekViewerFeedBack process 605 writes the concept 400) of the scored concept 603 Sto the dialogue history 604 and adds zero or more rated programs belonging to the concept 603 to the rating history 106. The seekViewerFeedback process 605 will be described in more detail below with reference to Fig. 8.

The getScoredConcepts process 600 as executed during the process 108, will now be described with reference to Fig. 7. The process 600 may be implemented as software resident in memory 1806 and being controlled in its execution by the processor 1805.

The process 600 begins by generating a concept power set 702 using a getConceptPowerSet process 701. The concept power set 702 represents the union of program description power sets associated with the program descriptions 103) in the EPG 100. A program description power set is a power set determined over the attributevalue pairs 201) of a program description 103). The program description power set is determined based on the attributes present in the salient attributes list 109.

The getScoredConcepts process 600 also scores 102) the program descriptions 103 within the EPG 100, using a scorePrograms process 703. The scorePrograms process 703 scores the program descriptions 103) within the EPG 100 using the program scorer 105 to produce a scored electronic program guide (EPG) 704.

For each concept 400) in the concept power set 702 the getScoredConcepts process 600 uses the scoreConcepts process 705 to determine the mean of the scores 102) of program descriptions 103) within the scored EPG 704 that the concept 690158.doc -24- O 400) covers. The output of the scoreConcepts process 705 is a scored concept power set ("1 o 601, which is the resulting power set of scored concepts 500).

As described above, the scored concept power set 601 is passed to the getDialogConcept process 602, which returns a scored concept 603, which is then passed 5 to the seekViewerFeedback process 605. The seekViewerFeedback process 605 will now 0\ 0 0 be described in detail below with reference to Fig. 8.

SThe seekViewerFeedback process 605 uses a getWatchedExemplars process 801 to Sselect a set of program descriptions 103) from the viewing history 104. The getWatchedExemplars process 801 randomly selects, from the viewing history 104, program descriptions 103) of television programs that have been watched by the user and that belong to the scored concept 603 the scored concept that was returned by the getDialogConcept process 602 of the process 108). The getWatchedExemplars process 801 creates a concept exemplar watched list 803 containing the selected program descriptions 103). The size of the concept exemplar watched list 803 may be set to five program descriptions 103).

The seekViewerFeedback process 605 also uses a getUnwatchedExemplars process 802 to select a set of program descriptions 103) from the EPG 100. The program descriptions selected by the getUnwatchedExemplars process 802 belong to the scored concept 603 that was returned by the getDialogConcept process 602 of the process 108.

The selected program descriptions also do not exist in the viewing history 104 as programs that have been watched by the user. The selected program descriptions also do not exist in the rating history 106. The selected program descriptions describe television programs that will be broadcast in the future and may be displayed by the system 1700.

The getUnwatchedExemplars process 802 creates a concept exemplar unwatched list 804 containing the selected program descriptions. The maximum size of the concept exemplar unwatched list 804 may be ten (10) program descriptions 103). Program 690158.doc descriptions with a shorter time until broadcast of the television program described by that program description may be added to the concept exemplar watched list 803 in preference to program descriptions describing television programs with a longer time until broadcast.

The seekViewerFeedback process 605 may use a userFeedbackDialog process 805 00to present the user with a feedback dialogue, via the TV 1712, for example. The r seekViewerFeedback process 605 may then add the concept 400) of the scored 0concept 603 the scored concept that was returned by the getDialogConcept process 602 of the process 108) to the dialog history 604.

An example of a feedback dialogue 900 will now be described with reference to Fig. 9. The feedback dialogue 900 contains a graphical representation 901 of a concept exemplar watched list 803. The graphical representation 901 is a list of the values of title attributes of program descriptions contained in the concept exemplar watched list 803 of Fig. 8.

The feedback dialogue 900 comprises a graphical representation 902 of a scored concept 603. The graphical representation 902 is a natural language description incorporating semantics of the attribute-value pairs of the concept 400) of the scored concept 603. The natural language description of the concept comprises the natural language description of the attribute-values that the concept contains, separated by "and".

The natural language description of an attribute-value of the concept is of the form "attribute is value" where the variable attribute is equal to the description of an attribute of the attribute-value, and value is the value of the attribute-value. An example, of a natural language description of an attribute-value pair is "Genre is comedy". The natural language description of a concept is formatted so that the description adheres to sentence case. Another example of a natural language description of the attribute-values of a concept is "Genre is comedy and rating is Different methods of producing natural 690158.doc -26language may be used. Linguistic knowledge specific to the preferred language of the o user of the system 1700 may be used to construct a natural language description of a concept 400). For example, English speaking people commonly use the term "action movies" to describe the concept that covers programs that are of the movie category, and of the action genre. Linguistic knowledge may be derived from examples 00 00 of usage of language as demonstrated by test subjects or written texts.

N The feedback dialogue 900 also comprises a graphical representation 903 of a 0concept exemplar unwatched list 804. The graphical representation 903 of a concept exemplar unwatched list takes the form of a graphical representation of the value of the title, channel and start time attributes of program descriptions 103) contained in the concept exemplar unwatched list 804 of Fig. 8.

The feedback dialogue 900 enables a user to select an element of the graphical representation 903 of the concept exemplar unwatched list 804. Such a selected element may be referred to as a "selected unwatched program". For example, the selected unwatched program 905 "Sliders") may be selected by the user. The feedback dialogue 900 displays a graphical representation of a program description 904 corresponding to the selected unwatched television program 905, on the TV 1712, for example. The graphical representation of a program description 904 may include a graphical representation of a title attribute-value 909, a genre attribute-value 910, a rating classification attribute-value 911, a content warnings attribute-value 912, a duration attribute-value 913, and a program synopsis attribute-value 914, for the selected unwatched television program 905.

The user may provide feedback as to the level of interest the user has in the selected unwatched television program 905, using the feedback dialogue 900. The user may select an interest value 907 from a range of interest values 906, using the directional buttons 1706 of the remote control unit 1705, for example. The range of interest values 906 may 690158.doc -27be the set of integers from one to five where a higher value represents a higher level of viewer interest.

The interest value 907 and the program description 103) that corresponds to e¢3 the selected unwatched television program 905 may be combined to form a viewer rating 302) that is added to the ratings history 106 configured within memory 1806 and/or 00 the hard disk drive 1810. Once the ratings history 106 is updated, the program scorer 105 Smay be reconfigured to incorporate the viewer rating as new user preference information.

The feedback dialogue 900 may enable the user of the system 1700 to flag the selected unwatched television program 905 for the purpose of indicating that the television program represented by the selected unwatched television program 905 should be recorded by the system 1700, using a video recorder (not shown) or DVD burner (not shown), for example. The selected unwatched television program 905 may be flagged using the remote control unit 1705, for example. The feedback dialogue 900 may comprise a graphical record program flag 908 representing whether a television program is flagged for recording. The feedback dialogue 900 may also comprise a means (not shown) to enable the user to flag the selected unwatched television program 905 to configure the system 1700 to generate a reminder five minutes before the start time of the selected unwatched television program 905.

In one implementation, a user may provide feedback on a level of interest in a concept 400), rather than a level of interest in television programs that belong to the concept.

Fig. 10 shows a rated concept 1000. The rated concept 1000 comprises a concept 1001 and a concept rating 1002. The concept rating 1002 comprises an integer value assigned by a user to the concept 1001. The integer value may be any one of the set of integers from one to five where a higher value indicates a higher level of interest by the user in the concept 1001.

690158.doc -28- Fig. 11 shows another example of a software architecture 1100 for implementing the described methods. In the architecture 1100, a rated concept program scorer 1103 returns a score 1102 as output when provided with a program description 103 as input.

C€3 The rated concept program scorer 1103 utilises a rated concepts list 1106, which comprises a list of rated concepts 1000) to determine the score 1102 for the program 00 00 description 103. A process 1900 for determining the score 1102 for the program (Ni Sdescription 103, as executed by the rated concept program scorer 1103, will now be described with reference to Fig. 19. The process 1900 may be implemented as software resident in memory 1806 and being controlled in its execution by the processor 1805.

The process 1900 begins at step 1901 where the processor 1805 forms a list of covering rated concepts consisting of all concepts within the rated concepts list 1100 that cover the program description 103. At the next step 1903, if the list of covering rated concepts is empty, the processor 1805 generates the score 1102 as a null score at the next step 1905. For example, the score 1102 may be set equal to negative one If the list of covering rated concepts is not empty, the description lengths of concepts within rated concepts contained in the list of covering rated concepts are observed, and a maximum concept description length is determined at step 1907. At the next step 1909, the rated concept program scorer 1103 generates the score 1102. At step 1909, the score 1102 is set to the average rating of rated concepts 1000) within the list of covering rated concepts that contain a concept with description length equal to the maximum concept description length.

The score 1102 may be set to the mode or median of rated concepts 1000) within the list of covering rated concepts that contain a concept with description length equal to the maximum concept description length.

As seen in Fig. 11, a recommender 1101 provides the program description 103 to the rated concept program scorer 1103, which returns the score 1102 for the program 690158.doc -29- O description 103. If the score 1102 is not a null score, the recommender 1101 may use the score 1102 as an indication of an expected level of interest of the user for the program description 103. If the score 1102 is a null score, the recommender 1101 provides the program description 103 to the program scorer 105, which returns the score 102. In this instance, the recommender 1101 may use the score 102 as an indication of an expected 00 0 0 level of interest of the user for the television program described by the program C description 103.

SThe architecture 1100 also comprises an updateRecommender process 1105. The updateRecommender process 1105 uses the getScoredConcepts process 600 and the getDialogConcept process 602, as described above. The updateRecommender process 1105 also uses a seekViewerFeedBack process 1200 to initiate a dialogue with the user.

As seen in Fig. 12, the seekViewerFeedback process 1200 uses the getWatchedExemplars process 801 to select a set of program descriptions 103) from the viewing history 104. As described above, the getWatchedExemplars process 801 selects program descriptions 103) from the viewing history 104 belonging to the scored concept 603 the scored concept that was returned by the getDialogConcept process 602 of the process 1105). The getWatchedExemplars process 801 creates the concept exemplar watched list 803 containing the selected program descriptions 103), as described with reference to Fig. 8.

The seekViewerFeedback process 1200 may use a userFeedbackDialog process 1201 to present the user with a feedback dialogue 1300, as seen in Fig. 13, via the TV 1712, for example. A rated concept 1104 is generated based on the concept 400) of the scored concept 603, and a concept rating 1002) based on the feedback from the feedback dialogue 1300. The rated concept 1104 is added to the rated concept list 1106, and the concept of the scored concept 603 is added to the dialog history 604.

690158.doc The feedback dialogue 1300 comprises the graphical representation of a concept Sexemplar watched list 901, which was described above with reference to Fig. 9. The d feedback dialogue 1300 also comprises the graphical representation 902 of a scored C€3 concept 500), which was described above with reference to Fig. 9.

In addition, the feedback dialogue 1300 comprises a feedback means 1301 to enable 00 the user of the system 1700 to provide feedback as to the level of interest the user has in r the concept corresponding to the graphical representation of a scored concept 902. The feedback means 1301 may comprise a dialogue question 1302, an affirmative response option 1303 and a negative response option 1304. The affirmative response option 1303 and the negative response option 1304 may be selected by the user using the remote control 1705. The dialogue question 1302 may ask the user whether the recommender 1101 should recommend television programs belonging to the concept corresponding to the graphical representation of a scored concept 902. In this instance, if the user selects the positive response option 1303 no action is taken. If the user selects the negative response option 1304, the rated concept 1104 of Fig. 11 is generated based on the concept of the scored concept 603 and a concept rating that is equal to the minimum value in the allowable range of rating values. The rated concept 1104 is then added to the rated concept list 1106 of Fig. 11.

Alternatively, the feedback means 1301 may enable the user to select a value from a range of rating values, where the values may be selected from the allowable range of rating values. In this instance, the rated concept 1104 may be generated comprising the concept of the scored concept 603, and a concept rating 1002) that is equal to the value selected by the user. The rated concept 1104 may then be added to the rated concept list 1106 ofFig. 11.

Fig. 14 shows a recommender search and configuration tool 1400. The recommender search and configuration tool 1400 may be implemented as software 690158.doc -31 resident in memory 1806 and being controlled in its execution by the processor 1805.

The recommender search and configuration tool 1400 receives as input the scored concept d power set 601, which is the output of the getScoredConcepts process 600 as described above with reference to Fig. 7. The recommender search and configuration tool 1400 also receives input from the rated concept list 1106, the EPG 100, and the viewing history 104.

00The recommender search and configuration tool (1400) may use a recommender search and configuration tool user interface 1500, as seen in Fig. 15, to display a list of scored concepts and rated concepts on the TV 1712, for example. The recommender (,i search and configuration tool user interface 1500 may be implemented as software resident in memory 1806 and being controlled in its execution by the processor 1805.

The recommender search and configuration tool user interface 1500 comprises a graphical representation 1501 of a concept list. The graphical representation 1501 of the concept list comprises natural language descriptions that use the semantics of the attribute-value pairs 401) of the concepts 400) of scored concepts 500) within the scored concept power set 601, and concepts 1001) of rated concepts 1000) within the rated concept list 1106. The form of the natural language description was described above with reference to Fig. 9. The graphical representation 1501 of a concept list may comprise ten (10) concepts 400) selected from the scored concept power set 601 and from the rated concept list 1106. The scored concepts 500) with high scores and rated concepts 1000) with high ratings may be selected for the graphical representation 1501 of a concept list in preference to those with low scores and ratings, respectively.

The recommender search and configuration tool user interface 1500 enables the user to select a selected concept 1502) from the graphical representation 1501, using the remote control unit 1705, for example.

690158.doc -32- Upon selecting the concept 1502, for example, the recommender search and o configuration tool user interface 1500 may display a graphical representation of a concept d exemplar unwatched list 1503 corresponding to the selected concept 1502. The graphical representation of a concept exemplar unwatched list 1503 is similar to the graphical representation 903, described above with reference to Fig. 9. The recommender search 00and configuration tool user interface 1500 also displays a graphical representation of a N' program description 1505 corresponding to a selected unwatched television program 1504, which is again similar to the graphical representation 904 described above with (,i reference to Fig. 9.

The recommender search and configuration tool user interface 1500 may also enable the user to provide feedback as to the level of interest the user has in the selected concept 1502 within the graphical representation of a concept list 1501. For example, as seen in Fig. 16, if the user selects a negative response option 1601 displayed on the TV 1712, using the remote control unit 1705, a rated concept 1000) may be generated based on the selected concept 1502 and a concept rating 1002) that is equal to a minimum value in an allowable range of rating values, where the allowable range of rating values may be equal to the allowable range of scores that may be output by the program scorer 105 of Fig 11. The rated concept may then be added to the rated concept list 1106 and the selected concept 1502 may be removed from within the graphical representation 1501, as seen in Fig. 16.

The aforementioned preferred method(s) comprise a particular control flow. There are many other variants of the preferred method(s) which use different control flows without departing the spirit or scope of the invention. Furthermore one or more of the steps of the preferred method(s) may be performed in parallel rather sequentially.

Industrial Applicability 690158.doc -33- It is apparent from the above that the arrangements described are applicable to the o computer and data processing industries.

The foregoing describes only some embodiments of the present invention, and modifications and/or changes can be made thereto without departing from the scope and spirit of the invention, the embodiments being illustrative and not restrictive.

00In the context of this specification, the word "comprising" means "including Ni, principally but not necessarily solely" or "having" or "including", and not "consisting only of'. Variations of the word "comprising", such as "comprise" and "comprises" have correspondingly varied meanings.

690158.doc

Claims (4)

1. A method of generating one or more scored concepts based on one or more data e¢3 items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted 00 00 level of interest in the associated data item, said method comprising the steps of: (Ni selecting one or more of the attributes associated with the data items; Sgenerating one or more concepts from the selected attributes, each of said concepts comprising at least one of the selected attributes; matching each of the generated concepts to one or more of the data items; statistically analysing the item scores of the data items matched to at least one of the generated concepts to generate a concept score for the at least one concept, wherein the generated concept score and associated concept represents a scored concept.

2. The method according to claim 1, wherein the statistical analysis determines the mean of the items scores of the matched data items.

3. The method according to claim 1, wherein the statistical analysis determines the mode of the items scores of the matched data items.

4. The method according to claim 1, wherein the statistical analysis determines the median of the item scores of the matched data items. A method of configuring a data item scorer, said data item scorer being adapted to generate an item score for each of one or more data items, the item score for an associated

690158.doc data item representing a predicted level of interest in the associated data item, said Smethod comprising the steps of: generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; 00 statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one generated concept, the generated concept score and associated concept representing a scored concept; selecting one of said data items based on said scored concept; displaying one or more attributes associated with the selected data item; inputting a rating for the selected data item; and configuring the data item scorer using the inputted rating. 6. The method according to claim 5, further comprising the steps of: adding the inputted rating to a history of ratings associated with one or more data items; and using the history of ratings to configure the data item scorer. 7. The method according to claim 5, further comprising the step of selecting one of said scored concepts. 8. The method according to claim 7, wherein the scored concept with the highest associated concept score is selected. 690158.doc -36- O 9. The method according to claim 7, wherein the selected concept has not been o previously selected. The method according to claim 5, wherein the inputted rating is selected from a range of ratings. 00 11. The method according to claim 5, wherein the selected data item is a television Oprogram. 12. The method according to claim 10, further comprising the step of flagging the selected television program for subsequent recording. 13. A user interface for configuring a data item scorer, said data item scorer being adapted to generate an item score for each of one or more data items, the item score for an associated data item representing a predicted level of interest in the associated data item, said user interface comprising: selection means for selecting one of said data items based on a scored concept; display means for displaying one or more attributes associated with the selected data item; and input means for inputting a rating for the selected data item, wherein one or more concepts are generated from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes, and the item scores of one or more of said data items that match at least one of the generated concepts are statistically analysed to generate said scored concept. 690158.doc 37 O 14. A method of determining a rated concept based on one or more data items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said method comprising the steps of: generating one or more concepts from one or more selected attributes associated 0with the data items, each of said concepts comprising at least one of the selected attributes; Ostatistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one generated concept, the generated concept score and associated concept representing a scored concept; displaying a natural language description of the scored concept; inputting a rating for the scored concept based on the displayed description; and associating the inputted rating with the scored concept to determine the rated concept. The method according to claim 14, the scored concept with the highest associated concept score is selected. 16. The method according to claim 14, further comprising the step of selecting one of said scored concepts. 17. The method according to claim 16, wherein the selected concept has not been previously selected. 690158.doc -38 O 18. The method according to claim 16, wherein the rated concept is adapted for o configuring a recommender of data items. 19. The method according to claim 14, wherein the inputted rating is selected from a range of ratings. 00 The method according to claim 14, wherein the data items are television programs. 21. The method according to. claim 20, further comprising the step of flagging a selected television program for subsequent recording. 22. A method of configuring a data item recommender using a rated concept based on a plurality of data items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said method comprising the steps of: generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one generated concept, the generated concept score and associated concept representing a scored concept; displaying a natural language description of the scored concept; 690158.doc -39 O inputting a rating for the scored concept based on the displayed description; o associating the inputted rating with the scored concept to determine the rated concept; and configuring the recommender using the rated concept. 0 23. A method of rating one or more data items for use in configuring a recommender of data items, each of the data items comprising one or more associated attributes values, said method comprising the steps of: selecting a concept from a set of concepts, each of said concepts comprising one or more of said data attribute values; selecting a data item based on the data attribute values of said concept; displaying one or more data attribute values associated with the selected data item; inputting a rating for the selected data item based on the displayed data attribute values for the data item; and storing the inputted rating and the selected data item for use in configuring the recommender. 24. A user interface for rating one or more data items for use in configuring a recommender of data items, said user interface comprising: first attribute display means for displaying one or more data attributes associated with a concept; data item display means for displaying one or more data items associated with the concept, the displayed data items being selectable and one or more of the displayed data items having been previously selected; second attribute display means for displaying one or more data attributes associated with at least one selected data item; and 690158.doc 40 O rating display means for displaying a rating associated with one or more of the data attributes displayed in the second data attribute display means, wherein the displayed rating is input using the user interface. 25. An apparatus for rating one or more data items for use in configuring a 0recommender of data items, each of the data items comprising one or more associated Nattributes values, said apparatus comprising: Oconcept selection means for selecting a concept from a set of concepts, each of said concepts comprising one or more of said data attribute values; data item selection means for selecting a data item based on the data attribute values of said concept; display means for displaying one or more data attribute values associated with the selected data item; input means for inputting a rating for the selected data item based on the displayed data attribute values for the data item; and storage means for storing the inputted rating and the selected data item for use in configuring the recommender. 26. A computer program for rating one or more data items for use in configuring a recommender of data items, each of the data items comprising one or more associated attributes values, said program comprising: code for selecting a concept from a set of concepts, each of said concepts comprising one or more of said data attribute values; code for selecting a data item based on the data attribute values of said concept; code for displaying one or more data attribute values associated with the selected data item; 690158.doc -41 O code for inputting a rating for the selected data item based on the displayed data attribute values for the data item; and code for storing the inputted rating and the selected data item for use in configuring the recommender. 0 27. An apparatus for generating one or more scored concepts based on one or more data (Ni items, each of the data items comprising one or more associated attributes and an Cassociated item score, the item score for an associated data item representing a predicted (,i level of interest in the associated data item, said apparatus comprising: selection means for selecting one or more of the attributes associated with the data items; concept generation means for generating one or more concepts from the selected attributes, each of said concepts comprising at least one of the selected attributes; matching means for matching each of the generated concepts to one or more of the data items; analysing means for statistically analysing the item scores of the data items matched to at least one of the generated concepts to generate a concept score for the at least one concept, wherein the generated concept score and associated concept represents a scored concept. 28. A computer program for generating one or more scored concepts based on one or more data items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said program comprising: code for selecting one or more of the attributes associated with the data items; 690158.doc 42 O code for generating one or more concepts from the selected attributes, each of said oconcepts comprising at least one of the selected attributes; code for matching each of the generated concepts to one or more of the data items; code for statistically analysing the item scores of the data items matched to at least one of the generated concepts to generate a concept score for the at least one concept, 0wherein the generated concept score and associated concept represents a scored concept. 29. An apparatus for configuring a data item scorer, said data item scorer being adapted to generate an item score for each of one or more data items, the item score for an associated data item representing a predicted level of interest in the associated data item, said apparatus comprising: concept generation means for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; analysing means for statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one generated concept, the generated concept score and associated concept representing a scored concept; selection means for selecting one of said data items based on said scored concept; display means for displaying one or more attributes associated with the selected data item; input means for inputting a rating for the selected data item; and configuring means for configuring the data item scorer using the inputted rating. 30. A computer program for configuring a data item scorer, said data item scorer being adapted to generate an item score for each of one or more data items, the item score for an 690158.doc 43 O associated data item representing a predicted level of interest in the associated data item, said program comprising: code for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; 0 code for statistically analysing the item scores of one or more of said data items that Nmatch at least one of the generated concepts to generate a concept score for the at least Oone generated concept, the generated concept score and associated concept representing a scored concept; code for selecting one of said data items based on said scored concept; code for displaying one or more attributes associated with the selected data item; code for inputting a rating for the selected data item based on the displayed data attribute values; and code for configuring the data item scorer using the inputted rating. 31. An apparatus for determining a rated concept based on one or more data items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said apparatus comprising: concept generation means for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; analysis means for statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one generated concept, the generated concept score and associated concept representing a scored concept; 690158.doc 44 O display means for displaying a natural language description of the scored concept; o input means for inputting a rating for the scored concept based on the displayed description; and association means for associating the inputted rating with the scored concept to determine the rated concept. 00 32. A computer program for determining a rated concept based on one or more data Oitems, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said program comprising: code for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; code for statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least one generated concept, the generated concept score and associated concept representing a scored concept; code for displaying a natural language description of the scored concept; code for inputting a rating for the scored concept based on the displayed description; and code for associating the inputted rating with the scored concept to determine the rated concept. 33. An apparatus for configuring a recommender using a rated concept based on a plurality of data items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item 690158.doc 45 O representing a predicted level of interest in the associated data item, said apparatus ocomprising: concept generation means for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; 0analysis means for statistically analysing the item scores of one or more of said data N items that match at least one of the generated concepts to generate a concept score for the Oat least one generated concept, the generated concept score and associated concept representing a scored concept; display means for displaying a natural language description of the scored concept; input means for inputting a rating for the scored concept based on the displayed description;associating the inputted rating with the scored concept to determine the rated concept; and configuration means for configuring the recommender using the rated concept. 34. A computer program for configuring a recommender using a rated concept based on a plurality of data items, each of the data items comprising one or more associated attributes and an associated item score, the item score for an associated data item representing a predicted level of interest in the associated data item, said program comprising: code for generating one or more concepts from one or more selected attributes associated with the data items, each of said concepts comprising at least one of the selected attributes; code for statistically analysing the item scores of one or more of said data items that match at least one of the generated concepts to generate a concept score for the at least 690158.doc 46 O one generated concept, the generated concept score and associated concept representing a scored concept; code for displaying a natural language description of the scored concept; code for inputting a rating for the scored concept based on the displayed description; 0code for associating the inputted rating with the scored concept to determine the N rated concept; and Ocode for configuring the recommender using the rated concept. 35. A method of generating one or more scored concepts based on a plurality of data items, said method being substantially as herein before described with reference to any one of the embodiments as that embodiment is shown in the accompanying drawings. 36. A method of configuring a data item scorer, said method being substantially as herein before described with reference to any one of the embodiments as that embodiment is shown in the accompanying drawings. 37. A user interface for configuring a data item scorer, said user interface being substantially as herein before described with reference to any one of the embodiments as that embodiment is shown in the accompanying drawings. 38. A method of determining a rated concept based on a plurality of data items, said method being substantially as herein before described with reference to any one of the embodiments as that embodiment is shown in the accompanying drawings. 690158.doc -47- 39. A method of configuring a recommender using a rated concept based on a plurality of data items, said method being substantially as herein before described with reference to any one of the embodiments as that embodiment is shown in the accompanying drawings. C€) 40. A method of rating one or more data items for use in configuring a recommender of 00 00 data items, said method being substantially as herein before described with reference to (N any one of the embodiments as that embodiment is shown in the accompanying drawings. 41. A method of rating one or more data items for use in configuring a recommender of data items, said method being substantially as herein before described with reference to any one of the embodiments as that embodiment is shown in the accompanying drawings. 42. A user interface for rating one or more data items for use in configuring a recommender of data items, said user interface being substantially as herein before described with reference to any one of the embodiments as that embodiment is shown in the accompanying drawings. DATED this Thirteenth Day of December 2004 CANON KABUSHIKI KAISHA Patent Attorneys for the Applicant SPRUSON&FERGUSON 690158.doc