Tracking changes in user-generated textual content on social media computing platforms

Patent metadata

Field	Value
Publication number	US-9817893-B2
Application number	US-201514625446-A
Country	US
Kind code	B2
Filing date	Feb 18, 2015
Priority date	Feb 18, 2015
Publication date	Nov 14, 2017
Grant date	Nov 14, 2017

Abstract

Social media posts related to a topic are analyzed over time by parsing the posts to identify terms and by statistically analyzing occurrences and co-occurrences of the terms in the posts to derive metrics. A relationship-based structure is updated over time based on the metrics. A relationship-based structure is updated over time based on the metrics. In an example, the relationship-based structure includes weighted nodes and edges. The nodes represent terms in the posts and the edges represent co-occurrences of the terms. The weights of the nodes depend on frequencies of the occurrences, while as the weights of the edges depend on frequencies of the co-occurrences. A trend in the social media posts is detected by identifying a change over time in the relationship-based data structure.

First claim

The invention claimed is:

1. In a network-based environment for analyzing social media data, a computer-implemented method performed on a computer for detecting a trend in social media posts on one or more social media platforms, the computer-implemented method comprising:
tracking occurrences of terms in the social media posts on the one or more social media platforms;
tracking co-occurrences of the terms in the social media posts by identifying terms that occur together in same social media posts;
updating a relationship-based data structure over time with metrics regarding the occurrences and co-occurrences, the relationship-based data structure comprising nodes and edges connecting the nodes, the nodes representing the terms, the edges representing the co-occurrences of the terms, the nodes and edges weighted based on the metrics, weights of the nodes being proportional to frequencies of the occurrences of the terms, and weights of the edges being proportional to frequencies of the co-occurrences;
identifying a first term of interest represented by a first node in the relationship-based data structure, the first term identified based at least in part on a weight of the first node;
identifying a story that comprises the first term of interest and a second term of interest, wherein the second term of interest is represented by a second node, and wherein the story is identified based on a weight of an edge that connects the first node and the second node;
detecting that the story is trending in the social media posts on the one or more social media platforms by identifying a change over time in the relationship-based data structure to the weight of the edge that connects the first node and the second node; and
providing a user interface that presents the story, wherein the user interface is updated over time based on the change to the weight of the edge, and wherein the updated user interface indicates that the story is trending.

2. The computer-implemented method of claim 1 , wherein identifying the first term of interest comprises:
selecting a set of the terms by comparing the weights of the nodes to a threshold; and
selecting the first term of interest from the set of the terms by comparing weights of edges associated with the set of the terms to another threshold.

3. The computer-implemented method of claim 1 , wherein detecting the that the story is trending comprises detecting an increase in the weight of the edge that connects the first node and the second node.

4. The computer-implemented method of claim 1 , further detecting that the story is losing interest of users of the one or more social media platforms based on the weight of the edge that connects the first node and the second node falling below a threshold.

5. The computer-implemented method of claim 1 , wherein detecting that the story is trending further comprises detecting an increase to the weight of first node representing.

6. The computer-implemented method of claim 1 , wherein a weight of a node representing a term is computed based on a frequency of occurrence of the term in the social media posts, a number of social media posts containing the terms, and a relevance of the social media posts containing the term.

7. The computer-implemented method of claim 1 , wherein a weight of an edge connecting two nodes corresponding to two terms is computed based on a frequency of the two terms occurring together in the social media posts and a relevance of a story that comprises the two terms, wherein the relevance is based on a frequency of the story occurring in the social media posts.

8. The computer-implemented method of claim 1 , wherein detecting that the story is trending comprises partitioning the relationship-based structure in a plurality of partitions, wherein nodes belonging to at least one partition represent trending terms of the story.

9. The computer-implemented method of claim 1 , wherein detecting that the story is trending comprises partitioning the relationship-based structure in a plurality of partitions by:
taking snapshots of the relationship-based structure at time intervals;
updating the relationship-based structure based on additional social media posts accessed between the time intervals;
computing frequencies of occurrences of the terms between the time intervals;
generating, as a function of the time intervals, similarities between the terms based on the frequencies of occurrences; and
assigning the nodes to the partitions based on the similarities between the terms.

10. The computer-implemented method of claim 1 , wherein detecting that the story is trending comprises partitioning the relationship-based structure in a plurality of partitions by:
taking snapshots of the relationship-based structure at time intervals;
updating the relationship-based structure based on additional social media posts accessed between the time intervals, the updating comprising increasing or decreasing weights of the edges based on co-occurrences of corresponding terms; and
assigning the nodes to the partitions based on the weights of the edges.

11. A computing system for analyzing social media posts regarding a story on one or more social media platforms, the computing system comprising:
a processor; and
a memory comprising computer-readable instructions that, when executed by the processor, cause the computing system to at least:
analyze occurrences of terms in the social media posts and co-occurrences of the terms in same social media posts to generate metrics associated with the terms over time;
update a graph data structure over time based on the metrics, the graph data structure comprising nodes representing the terms and edges connecting the nodes and representing co-occurrences of the terms, the nodes and edges being weighted based on the metrics, weights of the nodes being proportional to frequencies of the occurrences of the terms, and weights of the edges being proportional to frequencies of the co-occurrences;
select a set of the terms by comparing the weights of the nodes or the edges to a threshold, wherein the set of terms represents a set of terms of interest;
identify that the story is trending based on changes over time to the weights of nodes and edges corresponding to the set of terms, wherein the story comprises the set of terms of interest; and
provide a user interface that presents the story, wherein the user interface is updated over time based on the changes to the weights and wherein the updated user interface indicates that the story is trending.

12. The computing system of claim 11 , wherein identifying the trend comprises detecting the changes to the weight at time intervals, wherein the time intervals are defined as a function of a volume of the plurality of social media posts.

13. The computing system of claim 11 , wherein updates to the nodes and edges are tracked over time, wherein the trend of the story is identified based on the updates.

14. The computing system of claim 11 , wherein updates to the weights of the nodes and edges are tracked over time, wherein a start, an end, and a relevance of the story are identified based on the updates.

15. The computing system of claim 11 , wherein the social media posts comprises a plurality of posts received from the one or more social media platforms, wherein the updates are tracked by at least:
taking a snapshot of the graph data structure at a point in time;
receiving a post from a social media platform after the point in time;
analyzing occurrences and co-occurrences of terms of the post to determine changes to the weights of the nodes and

Assignees

• Adobe Systems Inc

Inventors

• Jaidka Kokil
• Kumaraguru Ponnurangam
• Chhaya Niyati
• Rustagi Sajal
• Gupta Prakhar
• Kaushik R

Classifications

• G06F16/345 Primary

  Summarisation for human users · CPC title

• G06F17/30719 Primary

  Physics · mapped topic

• H04L51/52

  for supporting social networking services · CPC title