Temporally sequenced content recommender method and system
A temporally sequenced content recommender method and system generates a first vector of affinities from usage information comprising durations of users' attention toward temporally sequenced content. A second vector of affinities is generated by applying a computer-implemented neural network to content objects. The similarity of the vectors of affinities is determined, which informs the generation of recommendations that are delivered to a user. Beneficial serendipity may be incorporated within the generation of the recommendations by, for example, evaluating contrasting affinities in user affinity vectors, by evaluating levels of available behavioral information for users, or by applying randomized or probabilistic methods. Explanations for the recommendations may be provided to users that include inferred user preferences.
1. A computer-implemented method, comprising:
accessing usage information comprising inferred durations of one or more users' focuses of attention that is associated with the one or more users' interactions with respect to each of a first plurality of content objects, wherein the each of the first plurality of content objects comprises temporally sequenced content;
generating automatically a first vector of affinities that is based upon the usage information;
generating automatically a plurality of affinity values by using a computer-implemented neural network that is applied to each of a second plurality of content objects, wherein the each of the second plurality of content objects comprises an element of each of a plurality of temporally sequenced content and each of the plurality of affinity values represents a degree of relationship that is associated with the each of the second plurality of content objects;
generating automatically a second vector of affinities, wherein the second vector of affinities comprises an arrangement of affinity values that includes at least one of the plurality of affinity values;
performing automatically a calculation of similarity between the first vector of affinities and the second vector of affinities;
determining automatically a recommendation of a content object comprising temporally sequenced content for delivery to a user of the one or more users, wherein the recommendation is determined in accordance with the calculation of similarity between the first vector of affinities and the second vector of affinities; and
delivering automatically the recommendation to the user.
2. The method of claim 1 , further comprising:
accessing the usage information comprising the inferred durations of the one or more users' focuses of attention associated with the one or more users' interactions with respect to the each of the first plurality of content objects, wherein the each of the first plurality of content objects comprise temporally sequenced images; and
determining the recommendation in response to a search request received from the user.
3. The method of claim 1 , further comprising:
accessing the usage information comprising the inferred durations of the one or more users' focuses of attention associated with the one or more users' interactions with respect to the each of the first plurality of content objects, wherein the each of the first plurality of content objects comprise temporally sequenced audio in the form of songs, and wherein the inferred durations comprise durations of the one or more users' listening.
4. The method of claim 1 , further comprising:
generating automatically the first vector of affinities, wherein each of the affinities are with respect to a distinct topical area and are derived from inferred relationship values between each pair of a plurality of users of the one or more users, wherein the inferred relationship values are determined by performing statistical learning.
5. The method of claim 1 , further comprising:
applying automatically a recommendation serendipity function, wherein the recommendation serendipity function incorporates beneficial serendipity in determining the recommendation of the content object.
6. The method of claim 5 , further comprising:
applying automatically the recommendation serendipity function in determining the recommendation of the content object, wherein the recommendation serendipity function identifies another vector of affinities that is determined to be similar to the first vector of affinities by applying a mathematical-based vector similarity algorithm and then identifying contrasting affinities between affinities of the first vector of affinities and affinities of the another vector of affinities that is based upon the usage information.
7. The method of claim 5 , further comprising:
applying automatically the recommendation serendipity function in determining the recommendation of the content object, wherein the recommendation serendipity function applies a probability distribution that is tunable.
8. The method of claim 1 , further comprising:
generating automatically a communication for delivery to the user that comprises an explanation for the recommendation, wherein the explanation is in a natural language format that comprises a plurality of syntactical elements that are each probabilistically selected and that refer to a theme derived from the second vector of affinities that is in accordance with an automatically inferred preference of the user.
9. A computer-implemented system comprising one or more processor-based devices configured to:
access usage information comprising inferred durations of one or more users' focuses of attention that is associated with the one or more users' interactions with respect to each of a first plurality of content objects, wherein the each of the first plurality of content objects comprises temporally sequenced content;
generate automatically a first vector of affinities that is based upon the usage information;
generate automatically a plurality of affinity values by using a computer-implemented neural network that is applied to each of a second plurality of content objects, wherein the each of the second plurality of content objects comprises an element of each of a plurality of temporally sequenced content and each of the plurality of affinity values represents a degree of relationship that is associated with the each of the second plurality of content objects;
generate automatically a second vector of affinities, wherein the second vector of affinities comprises an arrangement of affinity values that includes at least one of the plurality of affinity values;
perform automatically a calculation of similarity between the first vector of affinities and the second vector of affinities;
determine automatically a recommendation of a content object comprising temporally sequenced content for delivery to a user of the one or more users, wherein the recommendation is determined in accordance with the calculation of similarity between the first vector of affinities and the second vector of affinities; and
deliver automatically the recommendation to the user.
10. The system of claim 9 , further comprising the one or more processor-based devices configured to:
access the usage information comprising the inferred durations of the one or more users' focuses of attention associated with the one or more users' interactions with respect to the each of the first plurality of content objects, wherein each of the first plurality of content objects comprise temporally sequenced images; and
determining the recommendation in response to a search request received from the user.
11. The system of claim 9 , further comprising the one or more processor-based devices configured to:
access the usage information comprising the inferred durations of the one or more users' focuses of attention associated with the one or more users' interactions with respect to the each of the first plurality of content objects, wherein each of the first plurality of content objects comprise temporally sequenced audio in the form of songs, and wherein the inferred durations comprise durations of the one or more users' listening.
12. The system of claim 9 , further comprising the one or more processor-based devices configured to:
generate automatically the first vector of affinities, wherein each of the affinities are with respect to a distinct topical area and are derived from inferred relationship values between each pair of a plurality of users of the one or more users, wherein the inferred relationship values are determined by performing statistical learning.
13. The system of claim 9 , further comprising the one or more processor-based devices configured to:
apply automatically a recommendation serendipity function, wherein the recommendation serendipity function incorporates beneficial serendipity in determining the recommendation of the content object.
14. The system of claim 13 , further comprising the one or more processor-based devices configured to:
apply automatically the recommendation serendipity function in determining the recommendation of the content object, wherein the recommendation serendipity function identifies another vector of affinities that is determined to be similar to the first vector of affinities by applying a mathematical-based vector similarity algorithm and then identifying contrasting affinities between affinities of the first vector of affinities and affinities of the another vector of affinities that is based upon the usage information.
15. The system of claim 13 , further comprising the one or more processor-based devices configured to:
apply automatically the recommendation serendipity function in determining the recommendation of the content object, wherein the recommendation serendipity function applies a probability distribution that is tunable.
16. The system of claim 9 , further comprising the one or more processor-based devices configured to:
generate automatically a communication for delivery to the user that comprises an explanation for the recommendation, wherein the explanation is in a natural language format that comprises a plurality of syntactical elements that are each probabilistically selected and that refer to an automatically inferred preference of the user.
17. A computer-implemented system comprising one or more processor-based devices configured to:
access usage information comprising inferred durations of one or more users' focuses of attention that is associated with the one or more users' interactions with respect to each of a first plurality of content objects, wherein the each of the first plurality of content objects comprises temporally sequenced content;
generate automatically a first vector of affinities that is based upon the usage information;
generate automatically a plurality of affinity values by using a computer-implemented neural network that is applied to each of a second plurality of content objects, wherein the each of the second plurality of content objects comprises an element of each of a plurality of temporally sequenced content and each of the plurality of affinity values represents a degree of relationship that is associated with the each of the second plurality of content objects;
generate automatically a second vector of affinities, wherein the second vector of affinities comprises an arrangement of affinity values that includes at least one of the plurality of affinity values;
perform automatically a calculation of similarity between the first vector of affinities and the second vector of affinities;
apply automatically a recommendation serendipity function, wherein the recommendation serendipity function incorporates beneficial serendipity in a recommendation algorithm;
determine automatically by the recommendation algorithm a recommendation comprising a plurality of recommended content objects, wherein each of the plurality of recommended content objects comprises temporally sequenced content, wherein the recommendation is determined in accordance with the calculation of similarity between the first vector of affinities and the second vector of affinities and the application of the recommendation serendipity function; and
deliver automatically the recommendation to a user of the one or more users.
18. The system of claim 17 , further comprising the one or more processor-based devices configured to:
apply automatically the recommendation serendipity function, wherein the recommendation serendipity function identifies contrasting levels of available behavioral information between a pair of users of the one or more users, wherein the available behavioral information comprises inferred durations of the pair of users' focuses of attention that is associated with the pair of users' interactions with respect to temporally sequenced content.
19. The system of claim 17 , further comprising the one or more processor-based devices configured to:
apply automatically the recommendation serendipity function, wherein the recommendation serendipity function applies randomization.
20. The system of claim 17 , further comprising the one or more processor-based devices configured to:
generate automatically a communication for delivery to the user that comprises an explanation for the recommendation, wherein the explanation is in an automatically generated natural language format that refers to applying the serendipity and that is generated by probabilistically selecting each of a plurality of syntactical elements.