Method and system for topic disambiguation and classification
A method for generating recommendations involves selecting a first platform message, making a first determination that the first platform message is potentially associated with a plurality of topics including a first topic and a second topic, obtaining additional information associated with the first platform message including at least one of information about an account that authored the first platform message and information about third party accounts engaging with the first platform message, making a second determining that the first platform message is associated with the first topic using the plurality of topics and at least a portion of the additional information, wherein the first topic is an initial classification of the first platform message, generating a recommendation for at least one account based on the second determination, and providing the recommendation to at least one account.
1. A system comprising:
one or more computers and one or more storage devices on which are stored instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
selecting a message broadcasted on a social network platform, the message comprising content and having an author;
determining a content-based classification that classifies the message as related to two or more candidate topics of a plurality of topics, based on the content of the message, and for each of the candidate topics, generating a respective confidence score, wherein each confidence indicates a likelihood that the content is related to the respective candidate topic;
obtaining, from account profile information of the author on the social network platform, author topic scores for one or more of the candidate topics, wherein each author topic score is a topic interest score indicating a level of interest in, or a topic expertise score indicating a level of expertise about, the respective candidate topic;
generating an initial classification of the message from the content-based classification and the author topic scores, wherein the initial classification associates the message with one or more topics of the plurality of topics;
determining a final classification from the initial classification;
generating, based on the final classification, a recommendation for one or more other accounts of the social network platform; and
broadcasting the recommendation to the one or more other accounts.
2. The system of claim 1 , wherein:
the final classification is the initial classification; or
the final classification is determined after validating the initial classification, wherein validating the initial classification comprises:
obtaining topic engagement information about third party accounts engaging with the message, the topic engagement information comprising distributions of topic interest scores, topic expertise scores, or both, associated with the third party accounts; and
generating an engagement-based classification of the message based on the topic engagement information in addition to the content-based classification and the author topic scores and using the engagement-based classification in place of the initial classification as the final classification if the engagement-based classification associates different topics with the message.
3. The system of claim 1 , wherein:
the author topic scores include an expertise score for each of the candidate topics; and
generating the initial classification of the message uses only those author topic scores that are expertise scores.
4. The system of claim 1 , wherein:
the final classification associates the message with a single topic of the plurality of topics.
5. The system of claim 1 , wherein the operations comprise:
placing the message and an association of the message with the topics associated with the message by the final classification, in a message search pool in the social network platform.
6. The system of claim 5 , wherein the message search pool consists of messages broadcasted on the social network platform that are each associated with at least one of the plurality of topics.
7. The system of claim 6 , wherein the operations comprise:
receiving, by the social network platform, a search request for messages related to a particular topic; and
responding to the search request solely with messages associated with the particular topic in the message search pool.
8. One or more non-transitory computer storage media encoded with computer program instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:
selecting a message broadcasted on a social network platform, the message comprising content and having an author;
determining a content-based classification that classifies the message as related to two or more candidate topics of a plurality of topics, based on the content of the message, and for each of the candidate topics, generating a respective confidence score, wherein each confidence indicates a likelihood that the content is related to the respective candidate topic;
obtaining, from account profile information of the author on the social network platform, author topic scores for one or more of the candidate topics, wherein each author topic score is a topic interest score indicating a level of interest in, or a topic expertise score indicating a level of expertise about, the respective candidate topic;
generating an initial classification of the message from the content-based classification and the author topic scores, wherein the initial classification associates the message with one or more topics of the plurality of topics;
determining a final classification from the initial classification;
generating, based on the final classification, a recommendation for one or more other accounts of the social network platform; and
broadcasting the recommendation to the one or more other accounts.
9. The one or more non-transitory computer storage media of claim 8 , wherein:
the final classification is the initial classification; or
the final classification is determined after validating the initial classification, wherein validating the initial classification comprises:
obtaining topic engagement information about third party accounts engaging with the message, the topic engagement information comprising distributions of topic interest scores, topic expertise scores, or both, associated with the third party accounts; and
generating an engagement-based classification of the message based on the topic engagement information in addition to the content-based classification and the author topic scores and using the engagement-based classification in place of the initial classification as the final classification if the engagement-based classification associates different topics with the message.
10. The one or more non-transitory computer storage media of claim 8 , wherein:
the author topic scores include an expertise score for each of the candidate topics; and
generating the initial classification of the message uses only those author topic scores that are expertise scores.
11. The one or more non-transitory computer storage media of claim 8 , wherein:
the final classification associates the message with a single topic of the plurality of topics.
12. The one or more non-transitory computer storage media of claim 8 , wherein the operations comprise:
placing the message and an association of the message with the topics associated with the message by the final classification, in a message search pool in the social network platform.
13. The one or more non-transitory computer storage media of claim 12 , wherein the message search pool consists of messages broadcasted on the social network platform that are each associated with at least one of the plurality of topics.
14. The one or more non-transitory computer storage media of claim 13 , wherein the operations comprise:
receiving, by the social network platform, a search request for messages related to a particular topic; and
responding to the search request solely with messages associated with the particular topic in the message search pool.
15. A computer-implemented method comprising:
selecting a message broadcasted on a social network platform, the message comprising content and having an author;
determining a content-based classification that classifies the message as related to two or more candidate topics of a plurality of topics, based on the content of the message, and for each of the candidate topics, generating a respective confidence score, wherein each confidence indicates a likelihood that the content is related to the respective candidate topic;
obtaining, from account profile information of the author on the social network platform, author topic scores for one or more of the candidate topics, wherein each author topic score is a topic interest score indicating a level of interest in, or a topic expertise score indicating a level of expertise about, the respective candidate topic;
generating an initial classification of the message from the content-based classification and the author topic scores, wherein the initial classification associates the message with one or more topics of the plurality of topics;
determining a final classification from the initial classification;
generating, based on the final classification, a recommendation for one or more other accounts of the social network platform; and
broadcasting the recommendation to the one or more other accounts.
16. The computer-implemented method of claim 15 , wherein:
the final classification is the initial classification; or
the final classification is determined after validating the initial classification, wherein validating the initial classification comprises:
obtaining topic engagement information about third party accounts engaging with the message, the topic engagement information comprising distributions of topic interest scores, topic expertise scores, or both, associated with the third party accounts; and
generating an engagement-based classification of the message based on the topic engagement information in addition to the content-based classification and the author topic scores and using the engagement-based classification in place of the initial classification as the final classification if the engagement-based classification associates different topics with the message.
17. The computer-implemented method of claim 15 , wherein:
the author topic scores include an expertise score for each of the candidate topics; and
generating the initial classification of the message uses only those author topic scores that are expertise scores.
18. The computer-implemented method of claim 15 , wherein:
the final classification associates the message with a single topic of the plurality of topics.
19. The computer-implemented method of claim 15 , comprising:
placing the message and an association of the message with the topics associated with the message by the final classification, in a message search pool in the social network platform.
20. The computer-implemented method of claim 19 , wherein the message search pool consists of messages broadcasted on the social network platform that are each associated with at least one of the plurality of topics.
21. The computer-implemented method of claim 20 , comprising:
receiving, by the social network platform, a search request for messages related to a particular topic; and
responding to the search request solely with messages associated with the particular topic in the message search pool.