Systems and methods for generating alternative content recommendations
A computer-implemented method for obtaining one or more recommendation candidates for items of content available via a content distribution system, the method comprising: obtaining user data for a selected user, wherein the user data comprise or represent content metadata associated with user activity; performing a content recommendation process using the user data and content information for the available content to generate one or more initial content recommendation candidates for the user; generating or otherwise obtaining relationship information for the available content items and/or metadata associated with the available content items; generating one or more further content recommendation candidates from the initial content recommendation candidates using the relationship information.
1 . A computer-implemented method for obtaining one or more recommendation candidates for content items available via a content distribution system, the method comprising:
obtaining user data for a selected user from a storage resource, wherein the user data comprise or represent content metadata associated with user activity;
performing a content recommendation process using the user data and content information for the available content items to generate one or more initial content recommendation candidates for the user, in response to receiving a request from a user device;
generating or otherwise obtaining relationship information for the available content items and/or metadata associated with the available content items wherein the relationship information is based on at least content engagement data for a plurality of users of the content distribution system, wherein the relationship information comprises learned relationships between a plurality of available content items or learned relationships between a plurality of metadata items associated with the available content items and the plurality of available content items, wherein said relationship information is represented by a model or mapping between input content items or input metadata items and related output content items;
generating one or more further content recommendation candidates from the initial content recommendation candidates using the relationship information, wherein generating the one or more further content recommendation candidates comprises either a) or b):
a) applying the model or mapping to the initial content recommendation candidates to obtain the one or more further content recommendation candidates;
b) applying the model or mapping to metadata items associated with the initial content recommendation candidates to obtain the one or more further content recommendation candidates;
wherein the one or more further candidates match with the user data to a lesser degree than the initial one or more candidates.
2 . The method of claim 1 , wherein the one or more further content recommendation candidate candidates are different and/or at most overlap with the one or more initial content recommendation candidates generated using the user data.
3 . The method of claim 1 , wherein generating further content recommendation candidates using relationship information comprises applying a mapping from the one or more initial content recommendation candidates to the one or more further content recommendation candidates.
4 . The method of claim 1 , wherein the user data and/or the content information represent one or more properties comprise content parameters, properties and/or characteristics, such as programme title, time, duration, content type, programme categorisation, actor names, genre, release data, episode number, series number, style, mood, language and theme.
5 . The method of claim 1 , wherein generating the further content recommendation process is subject to one or more constraints such that the one or more further recommendation candidates comprise at least one candidate that has at least one property substantially different to content the user has previously engaged with.
6 . The method of claim 1 , comprising monitoring user activity including identifying content selected for viewing by the user of the user device and generating or updating the user data using metadata associated with the selected content.
7 . The method of claim 1 wherein the content information and/or user data represent and/or are indicative of metadata selected from a predetermined set of properties that comprises a stored ontology that includes at least 10,000 features that can be used as meta data to represent items of content, optionally wherein the ontology includes enriched versions of metadata obtained for items of content.
8 . The method of claim 1 , wherein the relationship information includes distance and/or separation information for the available content items and/or content metadata associated with the available content, and wherein the method comprises selecting a desired degree of distance or separation and generating the further content recommendation candidates based on the desired degree of distance or separation.
9 . The method of claim 1 , wherein the available content items and/or associated metadata are represented by discrete items and the relationship information is represented by a network between said discrete items, wherein the network comprises separation information, wherein the method comprises selecting a desired degree of separation when using the relationship information to generate the further recommendation candidates.
10 . The method of claim 1 , wherein the user data comprises or is based on user action data representing previous user actions, optionally content selection, viewing or recording actions, user language data and/or episode data and/or rating data and/or content metadata representing properties of content viewed, recorded or selected by a user.
11 . The method of claim 1 , wherein the relationship information comprises information on at least one relationship between the one or more content items and/or items of the associated metadata.
12 . The method of claim 1 , wherein the relationship information is represented as and/or form part of a first machine learning model and the content recommendation process is represented as and/or forms part of a second machine learning model such that the method comprises
applying the second machine learning model to at least part of the user data to generate the initial set of recommendation candidates;
applying the first machine learning model to at least one of the initial set of recommendation candidates to generate the plurality of content recommendation candidates.
13 . The method of claim 1 , wherein generating the further content recommendations using the relationship information comprises applying a mapping to the one or more of the initial content recommendations and/or their associated metadata, wherein the mapping is based on content engagement data for a plurality of users.
14 . The method of claim 1 , wherein a final set of recommendation candidates include at least some of the initial recommendation candidates and at least some of the further recommendation candidates wherein the method comprises controlling a weighting between the initial recommendation candidates and the further recommendation set of candidates.
15 . The method of claim 1 , wherein performing the content recommendation operation comprises receiving a request over a network for one or more recommendation candidates and sending the generated one or more recommendation candidates as a response over a network, wherein the request and response are received and sent using an predefined application protocol interface.
16 . A system comprising a user device, a storage resource and processing circuitry, wherein the processing circuitry is configured to:
obtain user data for a selected user from the storage resource, wherein the user data comprise or represent content metadata associated with user activity;
perform a content recommendation process using the user data and content information for available content items via a content distribution system to generate one or more initial content recommendation candidates for the user, in response to receiving a request from the user device;
generate or otherwise obtain relationship information for the available content items and/or metadata associated with the available content items wherein the relationship information is based on at least content engagement data for a plurality of users of the content distribution system, wherein the relationship information comprises learned relationships between a plurality of available content items or learned relationships between a plurality of metadata items associated with the available content items and the plurality of available content items, wherein said relationship information is represented by a model or mapping between input content items or input metadata items and related output content items;
generate further content recommendation candidates from the initial content recommendation candidates using the relationship information, wherein generating the one or more further content recommendation candidates comprises
generate one or more further content recommendation candidates from the initial content recommendation candidates using the relationship information, wherein generating the one or more further content recommendation candidates comprises either a) or b):
a) applying the model or mapping to the initial content recommendation candidates to obtain the one or more further content recommendation candidates;
b) applying the model or mapping to metadata items associated with the initial content recommendation candidates to obtain the one or more further content recommendation candidates;
wherein the one or more further candidates match with the user data to a lesser degree than the initial one or more candidates.
17 . A non-transitory computer-readable medium that comprises computer-readable instructions that are executable to perform a method according to claim 1 .
18 . The system of claim 1 , wherein at least one of a) and b):
the metadata that are common between the one or more further candidates and the initial one or more candidates have a lower associated weight for the one or more further candidates than for the initial one or more candidates.
19 . The system of claim 18 , wherein the one or more further candidates have more metadata that are not associated with the user activity than the initial one or more candidates.