Technique for building a user profile based on content consumption or production

• US 8,838,589 B1
• Filed: 08/16/2012
• Issued: 09/16/2014
• Est. Priority Date: 08/19/2011
• Status: Active Grant

• Current Assignee: RPX Corporation
• Original Assignee: Reverb Technologies, Inc.
• Inventors: Tam, Anthony L.
• Primary Examiner(s): Beausoliel, Jr., Robert
• Assistant Examiner(s): ALLEN, NICHOLAS E
• Application Number:
  US13/587,753
• Time in Patent Office: 761 Days
• Field of Search: 707/706, 707/722, 707/732, 707/737, 707/741, 707/748, 707/752, 707/754, 707/758, 707/769, 707/771, 707/780
• US Class Current: 707/732
• CPC Class Codes: G06F 16/337 Profile generation, learning or modification

Abstract

A method develops a user profile to indicate a user'"'"'s topics of interest. Documents considered by a computer system to have been consumed by first user are identified. Document summaries are developed identifying concepts represented in each document and relative strength by which the concept is considered associated with the document, and are accumulated from all the identified documents into a profile for the first user. The user profile is stored in a user profile database in a storage medium accessible to the computer system. A system develops a user profile to indicate the user'"'"'s topics of interest by identifying documents consumed by a user with a document summary for each document identifying concepts and their relative strength. Document summaries are accumulated into a profile for first user.

Claims

1. A method for developing a user profile for indicating the user'"'"'s topics of interest, comprising the steps of:

identifying a plurality of documents considered by a computer system to have been consumed by a first user;

for each identified document, a computer system developing a document summary identifying a plurality of concepts considered represented in the document and a relative strength by which the concept is considered associated with the document;

a computer system accumulating the document summaries from all the identified documents into a profile for the first user; and

storing the user profile in a user profile database in a storage medium accessible to the computer system,

wherein each of the plurality of concepts comprises, in combination, a normalized term used in the identified document and a usage attribute identifying a way in which the normalized term is used,

and wherein the step of developing a document summary, for a given identified document, comprises the steps of;

identifying a plurality of the concepts used in the identified document;

scoring a quality of each usage of the identified concepts in the identified document; and

developing, in dependence upon the quality scores for each given one of the identified concepts, a feature strength for the concept as used in the identified document,

the document summary indicating each of the concepts in the plurality of concepts in conjunction with its feature strength.

2. A method according to claim 1, wherein the step of identifying a plurality of documents considered by a computer system to have been consumed by a first user, comprises the step of determining, for each given one of a plurality of documents, a member of the group consisting of:

whether the first user has scrolled through the given document, and

whether the user remained on the given document for more than a predetermined minimum period of time and less than a predetermined maximum period of time.

3. A method according to claim 1, wherein the step of accumulating the document summaries from all the identified documents into a profile for the first user comprises the step of combining, separately for each of at least a plurality the concepts represented in the document summaries, the feature strength for the concept as used in each of the identified documents.

4. A method according to claim 1, wherein the step of accumulating the document summaries from all the identified documents into a profile for the first user comprises the step of combining, separately for each of at least a plurality the concepts represented in the document summaries, the relative strengths by which each concept is considered associated with the document.

5. A method according to claim 1, further comprising the step of storing in the user profile database a user profile for each of a plurality of users in addition to the first user.

6. A method according to claim 1, further comprising the step of searching a database for documents whose document summaries match a profile determined in dependence upon the user profile.

7. A method according to claim 1, further comprising the steps of:

blending, in accordance with predetermined weights, the user profile for the first user with a document summary for a document currently being consumed by the first user, to develop a blended user profile; and

searching a database for documents having document summaries which match a profile determined in dependence upon the blended user profile.

8. A system for developing a user profile for indicating the user'"'"'s topics of interest, the system comprising:

memory;

a data processor coupled to the memory, the data processor configured to;

identify a plurality of documents considered by the system to have been consumed by a first user;

for each identified document, develop a document summary identifying a plurality of concepts considered represented in the document and a relative strength by which the concept is considered associated with the document;

accumulate the document summaries from all the identified documents into a profile for the first user; and

store the user profile in a user profile database in a storage medium accessible to the system,

wherein each of the plurality of concepts comprises, in combination, a normalized term used in the identified document and a usage attribute identifying a way in which the normalized term is used,

and wherein in developing a document summary, for a given identified document, the data processor configured further to;

identify a plurality of the concepts used in the identified document;

score a quality of each usage of the identified concepts in the identified document; and

develop, in dependence upon the quality scores for each given one of the identified concepts, a feature strength for the concept as used in the identified document,

the document summary indicating each of the concepts in the plurality of concepts in conjunction with its feature strength.

9. A system according to claim 8, wherein the identifying a plurality of documents considered by a computer system to have been consumed by a first user, comprises determining, for each given one of a plurality of documents, a member of the group consisting of:

whether the first user has scrolled through the given document, and

whether the user remained on the given document for more than a predetermined minimum period of time and less than a predetermined maximum period of time.

10. A system according to claim 8, wherein accumulating the document summaries from all the identified documents into a profile for the first user comprises combining, separately for each of at least a plurality the concepts represented in the document summaries, the feature strength for the concept as used in each of the identified documents.

11. A system according to claim 8, wherein the data processor is further configured to:

blend, in accordance with predetermined weights, the user profile for the first user with a document summary for a document currently being consumed by the first user, to develop a blended user profile; and

search a database for documents having document summaries which match a profile determined in dependence upon the blended user profile.

12. A method for recommending content to a user, comprising:

storing in a database a user profile identifying a plurality of concepts considered of interest to the user, each in conjunction with a relative strength by which the concept is considered of interest to the user;

for a document currently being consumed by the user, developing a current document summary identifying a plurality of concepts considered represented in the document and a relative strength by which each concept is considered associated with the document;

blending the user profile with the current document summary to develop a blended user profile; and

searching a database for documents having document summaries which match a profile determined in dependence upon the blended user profile,

wherein the step of developing a current document summary comprises the steps of;

identifying a plurality of the concepts used in the document currently being consumed;

scoring a quality of each usage of the identified concepts in the document currently being consumed; and

developing, in dependence upon the quality scores for each given one of the identified concepts, a feature strength for the concept as used in the document currently being consumed,

the document summary indicating each of the concepts in the plurality of concepts in conjunction with its feature strength.

13. A method according to claim 12, wherein each of the concepts identified in either the user profile or the current document or both is identified by a normalized term in combination with a usage attribute identifying a way in which the term is used.

14. A method according to claim 12, wherein the step of blending the user profile with the current document summary comprises the step of, separately for each given one of a plurality of concepts included in both the user profile and the current document summary, averaging the relative strength identified for the given concept in the user profile with the relative strength identified for the given concept in the current document summary, each weighted by a respective blend factor.

15. A method according to claim 14, wherein the blend factor is the same for all concepts in the plurality of concepts included in both the user profile and the current document summary.

16. A method according to claim 12, wherein the step of blending the user profile with the current document summary comprises the step of, separately for each i′

th one of a plurality of concepts included in both the user profile and the current document summary, determining a blended feature strength given by a blend factor F times the relative strength identified for the given concept in the user profile, plus (1−

F) times the relative strength identified for the given concept in the current document summary.

17. A method for recommending content to a user, for use with a pool of candidate resources each having a respective document summary identifying a respective plurality of concepts considered represented in the candidate resource and a relative strength by which each concept is considered associated with the candidate resource, comprising the steps of:

storing in a database a user profile identifying a plurality of concepts considered of interest to the user, each in conjunction with a relative strength by which the concept is considered of interest to the user;

for a document currently being consumed by the user, developing a current document summary identifying a plurality of concepts considered represented in the document and a relative strength by which each concept is considered associated with the document;

blending the user profile with the current document summary to develop a blended user profile;

searching a database for documents having document summaries which match a profile determined in dependence upon the blended user profile;

pruning from the pool each given one of the candidate resources whose first n₁ concepts, in order of relative strength in the document summary for the given resource, does not match the first n₁ concepts identified in the blended user profile in order of relative strength in the blended user profile; and

presenting towards the user a list of resources in dependence upon the pruned pool,

where n₁ is a predetermined positive integer.

18. A method according to claim 17, wherein the step of presenting to the user a list of resources in dependence upon the pruned pool comprises the steps of:

further pruning from the pruned pool each particular one of the candidate resources whose next n₂ concepts, in order of relative strength in the document summary for the given resource, does not match the next n₂ concepts identified in the blended user profile in order of relative strength in the blended user profile; and

presenting towards the user a list of resources in dependence upon the further pruned pool,

where n₂ is a predetermined positive integer.

19. A method for recommending content to a user, for use with a pool of candidate resources each having a respective document summary identifying a respective plurality of concepts considered represented in the candidate resource and a relative strength by which each concept is considered associated with the candidate resource, comprising the steps of:

storing in a database a user profile identifying a plurality of concepts considered of interest to the user, each in conjunction with a relative strength by which the concept is considered of interest to the user;

for a document currently being consumed by the user, developing a current document summary identifying a plurality of concepts considered represented in the document and a relative strength by which each concept is considered associated with the document;

blending the user profile with the current document summary to develop a blended user profile;

pruning from the pool each given one of the candidate resources whose first n_i concepts, in order of relative strength in the document summary for the given resource, does not match the first n_i concepts identified in the blended user profile in order of relative strength in the blended user profile, where i=1;

for sequentially increasing integers i from i=2 until the pool identifies no more than a predetermined maximum number of resources, pruning from the pool each given one of the candidate resources whose next n_i concepts, in order of relative strength in the document summary for the given resource, does not match the next n_i concepts identified in the blended user profile in order of relative strength in the blended user profile; and

presenting towards the user a list of resources in dependence upon the pruned pool, where each n_i is a respective predetermined positive integer.

20. A method according to claim 19, wherein n_i>

n_i-1, for each i>

1.

21. A method according to claim 19, wherein each n_i=2ⁱ.

22. A system for recommending content to a user, for use with a pool of candidate resources each having a respective document summa identifying a respective plurality of concepts considered represented in the candidate resource and a relative strength by which each concept is considered associated with the candidate resource, the system comprising:

memory;

a data processor coupled to the memory, the data processor configured to;

store in a database a user profile identifying a plurality of concepts considered of interest to the user, each in conjunction with a relative strength by which the concept is considered of interest to the user;

for a document currently being consumed by the user, develop a current document summary identifying a plurality of concepts considered represented in the document and a relative strength by which each concept is considered associated with the document;

blend the user profile with the current document summary to develop a blended user profile;

search a database for documents having document summaries which match a profile determined in dependence upon the blended user profile,

prune from the pool each given one of the candidate resources whose first n₁ concepts, in order of relative strength in the document summary for the given resource, does not match the first n₁ concepts identified in the blended user profile in order of relative strength in the blended user profile; and

present towards the user a list of resources in dependence upon the pruned pool,

where n₁ is a predetermined positive integer.

23. A system according to claim 22, wherein each of the concepts identified in either the user profile or the current document or both is identified by a normalized term in combination with a usage attribute identifying a way in which the term is used.

24. A system according to claim 22, wherein developing a current document summary comprises:

identifying a plurality of the concepts used in the document currently being consumed;

scoring a quality of each usage of the identified concepts in the document currently being consumed; and

developing, in dependence upon the quality scores for each given one of the identified concepts, a feature strength for the concept as used in the document currently being consumed,

the document summary indicating each of the concepts in the plurality of concepts in conjunction with its feature strength.

25. A system according to claim 22, wherein blending the user profile with the current document summary comprises, separately for each given one of a plurality of concepts included in both the user profile and the current document summary, averaging the relative strength identified for the given concept in the user profile with the relative strength identified for the given concept in the current document summary, each weighted by a respective blend factor.

26. A system according to claim 25, wherein the blend factor is the same for all concepts in the plurality of concepts included in both the user profile and the current document summary.

27. A system according to claim 22, wherein blending the user profile with the current document summary comprises, separately for each i′

th one of a plurality of concepts included in both the user profile and the current document summary, determining a blended feature strength given by a blend factor F times the relative strength identified for the given concept in the user profile, plus (1−

F) times the relative strength identified for the given concept in the current document summary.