Generating News Headlines on Online Social Networks

Patent metadata

Field	Value
Publication number	US-2018143980-A1
Application number	US-201615359431-A
Country	US
Kind code	A1
Filing date	Nov 22, 2016
Priority date	Nov 22, 2016
Publication date	May 24, 2018
Grant date	—

Abstract

In one embodiment, a method includes identifying a trending topic on an online social network, accessing a plurality of content objects posted to the online social network, wherein each content object is associated with the trending topic, and categorizing each content object into clusters based on a natural-language analysis of the content objects. The method may further include calculating a quality score for each cluster, wherein the quality score for each cluster is based at least on a measure of recency of one or more publication dates of the content objects within the cluster, select the cluster with the highest quality score as a trending cluster, and generating a trending-topic interface that includes a headline and description of the trending topic, wherein the headline and description are extracted from one or more of the content objects within the trending cluster.

First claim

What is claimed is:

1 . A method comprising:
identifying a trending topic on an online social network; accessing a plurality of content objects posted to the online social network, wherein each content object is associated with the trending topic; categorizing each content object into one of a plurality of discrete clusters based on a natural-language analysis of the respective content objects, wherein the trending topic is associated with a plurality of sub-topics, and wherein each cluster relates to a different sub-topic of the plurality of sub-topics; calculating a quality score for each cluster, wherein the quality score for each cluster is based at least on a measure of recency of one or more publication dates of the content objects within the cluster; selecting the cluster with the highest quality score as a trending cluster; and generating a trending-topic interface comprising a headline and description of the trending topic, wherein the headline and description are extracted from one or more of the content objects within the trending cluster.

2 . The method of claim 1 , further comprising sending the trending-topic interface to one or more client systems associated with one or more users of the online social network, respectively.

3 . The method of claim 1 , wherein the headline and description of the trending topic appears in a context module for the trending topic.

4 . The method of claim 1 , further comprising:
extracting one or more text strings selected from one or more content objects within the trending cluster; and generating the headline and description of the trending topic based on the extracted text strings.

5 . The method of claim 4 , wherein the extracted text strings have appeared in a number of user-generated posts above a threshold number of user-generated posts.

6 . The method of claim 4 , wherein the content objects comprise one or more articles posted on the online social network, and wherein the extracted text strings are headlines from the articles.

7 . The method of claim 6 , wherein generating the headline of the trending topic comprises:
comparing, for each extracted headline from each article of the one or more articles, the n-grams of the extracted headline with the n-grams of the extracted headline from each other article of the one or more articles; and selecting one of the extracted headlines as the headline of the trending topic based on the selected headline having more words in common with the extracted headlines from each other article than any of the extracted headlines from each other article have in common with each other.

8 . The method of claim 1 , wherein each sub-topic is associated with an event that has occurred within a threshold timeframe.

9 . The method of claim 1 , wherein the quality score for each cluster is further based on a number of content objects within the cluster.

10 . The method of claim 1 , wherein the quality score for each cluster is further based on a measure of coherence of content objects within the cluster.

11 . The method of claim 1 , wherein the quality score for each cluster is further based on a measure of relevance of the content objects within the cluster to the trending topic, wherein the measure of relevance is calculated by a vector-space model that measures a similarity between a vector representation of each content object in the cluster and vector representation of the trending topic.

12 . The method of claim 1 , wherein the quality score for each cluster is further based on a proportion of content objects within the cluster that were published by verified publishers.

13 . The method of claim 1 , wherein the plurality of content objects comprises one or more articles.

14 . The method of claim 1 , wherein the plurality of content objects comprises one or more user-generated posts.

15 . The method of claim 1 , wherein the natural-language analysis is a term frequency-inverse document frequency (TF-IDF) analysis.

16 . The method of claim 1 , wherein the trending-topic interface further comprises one or more images associated with the trending topic, wherein the one or more images are extracted from one or more of the content objects within the trending cluster.

17 . One or more computer-readable non-transitory storage media embodying software that is operable when executed to:
identify a trending topic on an online social network; access a plurality of content objects posted to the online social network, wherein each content object is associated with the trending topic; categorize each content object into one of a plurality of discrete clusters based on a natural-language analysis of the respective content objects, wherein the trending topic is associated with a plurality of sub-topics, and wherein each cluster relates to a different sub-topic of the plurality of sub-topics; calculate a quality score for each cluster, wherein the quality score for each cluster is based at least on a measure of recency of one or more publication dates of the content objects within the cluster; select the cluster with the highest quality score as a trending cluster; and generate a trending-topic interface comprising a headline and description of the trending topic, wherein the headline and description are extracted from one or more of the content objects within the trending cluster.

18 . A system comprising: one or more processors; and a non-transitory memory coupled to the processors comprising instructions executable by the processors, the processors operable when executing the instructions to:
identify a trending topic on an online social network; access a plurality of content objects posted to the online social network, wherein each content object is associated with the trending topic; categorize each content object into one of a plurality of discrete clusters based on a natural-language analysis of the respective content objects, wherein the trending topic is associated with a plurality of sub-topics, and wherein each cluster relates to a different sub-topic of the plurality of sub-topics; calculate a quality score for each cluster, wherein the quality score for each cluster is based at least on a measure of recency of one or more publication dates of the content objects within the cluster; select the cluster with the highest quality score as a trending cluster; and generate a trending-topic interface comprising a headline and description of the trending topic, wherein the headline and description are extracted from one or more of the content objects within the trending cluster.

Assignees

• Facebook Inc

Inventors

• Tanikella Soorya Vamsi Mohan
• Mashiach Li-Tal
• Mazepov Ivan
• Sinha Apurva

Classifications

• G06F16/24578 Primary

  using ranking · CPC title

• G06F16/285

  Clustering or classification · CPC title

• G06Q10/10

  Office automation; Time management · CPC title

• G06F16/9535

  Search customisation based on user profiles and personalisation · CPC title

• G06Q30/0241

  Advertisements · CPC title