Search relevance using messages of a messaging platform

Patent metadata

Field	Value
Publication number	US-10242003-B2
Application number	US-201715832751-A
Country	US
Kind code	B2
Filing date	Dec 5, 2017
Priority date	Mar 5, 2013
Publication date	Mar 26, 2019
Grant date	Mar 26, 2019

Abstract

A method and system for searching documents. The method can include: identifying a first set of messages of a messaging platform referencing a document; identifying message text of the first set of messages; associating the message text with the document in a storage repository; receiving, by a computer processor, a search request including a search term provided by a client; calculating, by the computer processor, a relevance score for the document based at least on the search term and the message text; and providing, to the client and based on the relevance score, a reference to the document in response to the search request.

First claim

What is claimed is:

1. A method for searching documents, comprising:
identifying a first set of messages of a messaging platform referencing a document comprising a Web page;
identifying message text of the first set of messages;
associating the message text with the document in a storage repository;
receiving, by a computer processor, a search request comprising a search term provided by a client;
calculating, by the computer processor, a relevance score for the document based at least on the search term and the message text;
providing, to the client and based on the relevance score, a reference to the document in response to the search request;
identifying a second set of messages posted to the messaging platform and referencing another document comprising a second web page;
analyzing message text of the second set of messages;
determining, based on a predefined relevance signal, that the first web page is relevant to the second web page; and
calculating a second relevance score for the second web page based at least on the search term and the message text of the second set of messages, wherein calculating the relevance score is further based on the second relevance score.

2. The method of claim 1 , further comprising calculating, based on the message text and the search term, a message relevance score for each message of the first set of messages; and calculating a representative message relevance score based on the message relevance scores, wherein calculating the relevance score for the document is further based on the representative message relevance score.

3. The method of claim 2 , further comprising: identifying, for each message of the first set of messages, an engagement value representing user engagement with the message, wherein the message relevance score of the message is further based on the engagement value.

4. The method of claim 1 , further comprising: identifying a set of conversations associated with the first set of messages; identifying a set of reply messages in the set of conversations; and analyzing message text of the set of reply messages, wherein calculating the relevance score is further based on the message text of the set of reply messages.

5. The method of claim 4 , further comprising: identifying a maximum threshold degree of separation value (N); and determining that each of the set of reply messages is within N degrees of separation of at least one message of the first set of messages.

6. The method of claim 4 , further comprising: determining that a first subset of the set of reply messages are each within one degree of separation of at least one message of the first set of messages; and determining that a second subset of the set of reply messages are each within two degrees of separation of at least one message of the first set of messages, wherein message text of the first subset is weighted more heavily than message text of the second subset when calculating the relevance score.

7. The method of claim 1 , wherein the predefined relevance signal is a common base uniform resource locator (URL) of the first web page and the second web page.

8. The method of claim 1 , further comprising:
calculating an engagement value associated with the first set of messages, wherein calculating the relevance score is further based on the engagement value.

9. The method of claim 8 , wherein the engagement value is one selected from a group consisting of an average number of message rebroadcasts, an average number of favorites, an average number of likes, and an average number of replies.

10. The method of claim 1 , further comprising:
extracting document text from the document;
analyzing the document text to determine that at least one relevant term of the message text is not found in the document text; and
associating the document text with the document in the storage repository, wherein calculating the relevance score is further based on the document text, and wherein the relevant term is weighed more heavily in calculating the relevance score than terms of the message text which also reside in the document text.

11. A system for searching documents, comprising:
a computer processor;
an aggregation module executing on the computer processor and configured to:
identify a first set of messages of a messaging platform referencing a document comprising a Web page; identify message text of the first set of messages; and
associate the message text with the document in a storage repository; and

a search module configured to:
receive a search request comprising a search term provided by a client;
calculate a relevance score for the document based at least on the search term and the message text;
provide, to the client and based on the relevance score, a reference to the document in response to the search request;
identify a second set of messages posted to the messaging platform and referencing another document comprising a second web page;
analyze message text of the second set of messages;
determine, based on a predefined relevance signal, that the first web page is relevant to the second web page; and
calculate a second relevance score for the second web page based at least on the search term and the message text of the second set of messages, wherein calculating the relevance score is further based on the second relevance score.

12. The system of claim 11 , wherein the aggregation module is further configured to:
calculate, based on the message text and the search term, a message relevance score for each message of the first set of messages; and
calculate a representative message relevance score based on the message relevance scores, wherein calculating the relevance score for the document is further based on the representative message relevance score.

13. The system of claim 12 , wherein the aggregation module is further configured to:
identify, for each message of the first set of messages, an engagement value representing user engagement with the message, wherein the message relevance score of the message is further based on the engagement value.

14. The system of claim 11 , wherein the aggregation module is further configured to:
identify a set of conversations associated with the first set of messages;
identify a set of reply messages in the set of conversations; and
analyze message text of the set of reply messages, wherein calculating the relevance score is further based on the message text of the set of reply messages.

15. The system of claim 14 , wherein the aggregation module is further configured to:
identify a maximum threshold degree of separation value (N); and
determine that each of the set of reply messages is within N degrees of separation of at least one message of the first set of messages.

16. The system of claim 14 , wherein the aggregation module is further configured to:
determine that a first subset of the set of reply messages are each within one degree of separation of at least one message of the first set of messages; and
determine that a second subset of the set of reply messages are each within two degrees of separation of at least one message of the first set of messages, wherein message text of the first subset is weighted more heavily than message text of the second subset when calculating the relevance score.

17. A non-transitory computer-readable storage medium comprising a plurality of instructions for searching documents, the plurality of instructions executing on at least one computer processor and comprising functionality to:
identify a first set of messages of a mess

Assignees

• Twitter Inc

Inventors

• Mishne Gilad
• Lin Jimmy

Classifications

• G06F16/93 Primary

  Document management systems · CPC title

• G06F16/313

  Selection or weighting of terms for indexing · CPC title

• G06F17/30011 Primary

  Physics · mapped topic

• G06F17/30616

  Physics · mapped topic