Systems and methods for ranking comments

• US 10,630,632 B2
• Filed: 11/12/2015
• Issued: 04/21/2020
• Est. Priority Date: 11/12/2015
• Status: Active Grant

• Current Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Original Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Inventors: Owens, Erich James
• Primary Examiner(s): Hu, Jensen
• Application Number:
  US14/939,827
• Publication Number:
  US 20170140036A1
• Time in Patent Office: 1,622 Days
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: H04L 51/52 for supporting social networking services

Abstract

Systems, methods, and non-transitory computer-readable media can train a model to define relatedness ratings for a plurality of terms. A posted content item and a comment to the posted content item are received. A relevance rating for the comment and the posted content item is determined based on the model. The comment is ranked among a plurality of comments based on the relevance rating.

Claims

1. A computer-implemented method comprising:

training, by a computing system, a model to define relatedness ratings for a plurality of terms, each relatedness rating indicative of a relatedness between two terms;

receiving, by the computing system, a posted content item;

receiving, by the computing system, a comment posted by a user to the posted content item;

determining, by the computing system, a relevance rating for the comment and the posted content item based on the model; and

ranking, by the computing system, the comment among a plurality of comments based on the relevance rating and further based on comment interaction information indicative of interaction with the comment by users of a social networking system, wherein

ranking the comment among the plurality of comments based on the comment interaction information comprises ranking the comment among the plurality of comments based on an interaction-to-impression ratio associated with the comment and user information associated with the users, wherein the interaction-to-impression ratio is determined based in part on a user sample size and the user information comprises posting history associated with the users.

2. The computer-implemented method of claim 1, further comprising defining a multi-dimensional space based on the model.

3. The computer-implemented method of claim 2, further comprising mapping the posted content item and the comment in the multi-dimensional space.

4. The computer-implemented method of claim 3, wherein determining a relevance rating for the comment and the posted content item comprises calculating a cosine similarity between the posted content item and the comment in the multi-dimensional space.

5. The computer-implemented method of claim 1, wherein training the model to define relatedness ratings for a plurality of terms comprises analyzing a plurality of pages on a social networking system for related terms.

6. The computer-implemented method of claim 1, wherein ranking the comment among a plurality of comments based on the relevance rating comprises determining whether the relevance rating satisfies a lower relevance rating threshold.

7. The computer-implemented method of claim 1, wherein ranking the comment among a plurality of comments based on the relevance rating comprises determining whether the relevance rating satisfies an upper relevance rating threshold.

8. The computer-implemented method of claim 7, wherein no ranking benefit is conferred by the relevance rating if the relevance rating exceeds the upper relevance rating threshold.

9. The computer-implemented method of claim 1, wherein ranking the comment among a plurality of comments based on the relevance rating comprises crediting an interaction rating based on the relevance rating.

10. The computer-implemented method of claim 1, wherein receiving a posted content item comprises analyzing a non-text posted content item and associating one or more terms with the non-text posted content item.

11. A system comprising:

at least one processor; and

a memory storing instructions that, when executed by the at least one processor, cause the system to perform;

training a model to define relatedness ratings for a plurality of terms, each relatedness rating indicative of a relatedness between two terms;

receiving a posted content item;

receiving a comment posted by a user to the posted content item;

determining a relevance rating for the comment and the posted content item based on the model; and

ranking the comment among a plurality of comments based on the relevance rating and further based on comment interaction information indicative of interaction with the comment by users of a social networking system, wherein

ranking the comment among the plurality of comments based on the comment interaction information comprises ranking the comment among the plurality of comments based on an interaction-to-impression ratio associated with the comment and user information associated with the users, wherein the interaction-to-impression ratio is determined based in part on a user sample size and the user information comprises posting history associated with the users.

12. The system of claim 11, wherein the method further comprises defining a multi-dimensional space based on the model.

13. The system of claim 12, wherein the method further comprises mapping the posted content item and the comment in the multi-dimensional space.

14. The system of claim 13, wherein determining a relevance rating for the comment and the posted content item comprises calculating a cosine similarity between the posted content item and the comment in the multi-dimensional space.

15. The system of claim 11, wherein training the model to define relatedness ratings for a plurality of terms comprises analyzing a plurality of pages on a social networking system for related terms.

16. A non-transitory computer-readable storage medium including instructions that, when executed by at least one processor of a computing system, cause the computing system to perform a method comprising:

training a model to define relatedness ratings for a plurality of terms, each relatedness rating indicative of a relatedness between two terms;

receiving a posted content item;

receiving a comment by a user to the posted content item;

determining a relevance rating for the comment and the posted content item based on the model; and

ranking the comment among a plurality of comments based on the relevance rating and further based on comment interaction information indicative of interaction with the comment by users of a social networking system, wherein

ranking the comment among the plurality of comments based on the comment interaction information comprises ranking the comment among the plurality of comments based on an interaction-to-impression ratio associated with the comment and user information associated with the users, wherein the interaction-to-impression ratio is determined based in part on a user sample size and the user information comprises posting history associated with the users.

17. The non-transitory computer-readable storage medium of claim 16, wherein the method further comprises defining a multi-dimensional space based on the model.

18. The non-transitory computer-readable storage medium of claim 17, wherein the method further comprises mapping the posted content item and the comment in the multi-dimensional space.

19. The non-transitory computer-readable storage medium of claim 18, wherein determining a relevance rating for the comment and the posted content item comprises calculating a cosine similarity between the posted content item and the comment in the multi-dimensional space.

20. The non-transitory computer-readable storage medium of claim 16, wherein training the model to define relatedness ratings for a plurality of terms comprises analyzing a plurality of pages on a social networking system for related terms.