Clustering of documents for providing content

A system for selecting content for a device identifier comprising a memory and one or more processors, wherein the memory stores instructions that when executed by the one or more processors, perform operations to:

• receive a post to a social network associated with the device identifier, the post having one or more first keywords;
• retrieve, from a source database attributed to one or more interest identifiers, a labeled document including one or more entries, each of the one or more entries of the labeled document including one or more second keywords, each of the one or more entries tagged with the one or more interest identifiers, each of the one or more entries comprising a previous post stored at the source database;
• extract the one or more first keywords of the post;
• extract the one or more second keywords for each of the one or more entries of the labeled document;
• generate a first relatedness metric between a first subset of the one or more first keywords of the post with a first subset of the one or more entries of the labeled document;
• generate a second relatedness metric between a second subset of the one or more first keywords of the post with a second subset of the one or more entries of the labeled document;
• determine that a difference between the first relatedness metric and the second relatedness metric is less than a predetermined threshold;
• generate, responsive to determining that the difference is less than the predetermined threshold, a post-entry cluster by associating the second subset of the one or more first keywords of the post with the second subset of the one or more entries of the labeled document based on the second relatedness metric between the second subset of one or more first keywords and the second subset of the one or more second keywords, each first keyword of the second subset of the one or more first keywords tagged with the interest identifier of the one or more entries associated with the second subset of the one or more second keywords;
• associate the post with the one or more entries of the labeled document based, at least in part, on the post-entry cluster;
• determine a prevalent interest identifier of the post based in part on a number of the on more interest identifiers of the one or more entries associated with the second subset of the one or more first keywords in the post-entry cluster;
• add the post to the one or more entries of the labeled document for generating a second post-entry cluster, the post tagged with the prevalent interest identifier;
• associate the prevalent interest identifier with the device identifier;
• and
• select content for the device identifier based, in part, on the prevalent interest identifier of the post.

A computer-implemented method for associating interests with device identifiers, comprising:

• receiving, by one or more processing systems, a plurality of posts from a social network associated with a device identifier;
• retrieving, by the one or more processing systems, from a source database attributed to one or more interest identifiers, a labeled document including a plurality of entries, each entry associated with the one or more interest identifiers, each of the one or more entries comprising a previous post stored at the source database;
• extracting, by the one or more processing systems, one or more first keywords from each post of the plurality of posts and one or more second keywords from each entry of the plurality of entries;
• generating, by the one or more processing systems, a first relatedness metric between a first subset of the one or more first keywords of the plurality of posts with a first subset of the one or more entries of the labeled document;
• generating, by the one or more processing systems, a second relatedness metric between a second subset of the one or more first keywords of the plurality of posts with a second subset of the one or more entries of the labeled document;
• determining, by the one or more processing systems, that a difference between the first relatedness metric and the second relatedness metric is less than a predetermined threshold;
• generating, by the one or more processing systems, responsive to determining that the difference is less than the predetermined threshold, a plurality of post-entry clusters, each post-entry cluster associating the second subset of the one or more first keywords of the respective post with one or more entries of the labeled document based on the second relatedness metric between the second subset of the one or more keywords and the second subset of the one or more second keywords, each first keyword of the second subset of the one or more first keywords tagged with the interest identifier of the one or more entries associated with the second subset of the one or more second keywords;
• associating, by the one or more processing systems, each post with the one or more entries of the labeled document based in part on the plurality of post-entry clusters;
• determining, by the one or more processing systems, a prevalent interest identifier of the plurality of posts based in part on a number of the one or more interest identifiers of the one or more entries associated with the second subset of the one or more first keywords in the plurality of post-entry clusters;
• adding, by the one or more processing systems, the plurality of posts to the plurality of entries of the labeled document for generating a second post-entry cluster, each of the plurality of posts tagged with the prevalent interest;
• associating, by the one or more processing systems, the prevalent interest identifier with the device identifier;
• receiving, by the one or more processing systems, a request for content from a client device associated with the device identifier;
• and
• transmitting, by the one or more processing systems, data to effect display of the content on the client device, wherein the content is selected based, in part, on the prevalent interest identifier associated with the device identifier.