Generating and using socially-curated brains

Inventors

Jakubik, Paul A. • Hagar, David Adam • Copps, David S.

Assignees

Brainspace Corp

Abstract

A method includes determining a plurality of social objects, each social object having a link to a link object on a network. The method further includes applying a filter to the determined social objects in order to determine a plurality of filtered social objects, retrieving a copy of each of the link objects linked to by the plurality of filtered social objects, and generating, using the retrieved copies of the link objects linked to by the plurality of filtered social objects, a matrix comprising a plurality of vectors. The method further includes generating a singular value representation of the matrix by performing Singular Value Decomposition (SVD) on the matrix and storing the singular value representation of the matrix in one or more memory units.

Core Innovation

The invention addresses network search by using socially-curated brain representations derived from social objects. Social objects include tweets or messages that link to link objects on a network such as the Internet or a corporate network. From stored social objects, a first group of tweets is determined and then filtered to obtain a plurality of filtered tweets.

For the filtered tweets, copies of the linked link objects are retrieved. The retrieved copies are used to generate a matrix comprising a plurality of vectors. The matrix is converted into a singular value representation by performing Singular Value Decomposition on the matrix.

The singular value representation is stored in one or more memory units and is then used to perform a query using at least a portion of the singular value representation. The filtering is based on selecting tweets or social objects whose links have been linked to a predetermined number of times and that have been retweeted or re-posted a predetermined number of times or by a predetermined number of users.

Claims Coverage

The independent claims define three core claim sets that share the same inventive pipeline and each include multiple inventive features. Across the independent claims, there are five main inventive features: determining social objects with links, filtering those social objects based on engagement and link popularity thresholds, retrieving linked objects, generating a matrix and computing a singular value representation using Singular Value Decomposition, and storing the representation for query processing.

Each independent claim centers on building a singular value representation from a matrix generated from retrieved copies of link objects referenced by filtered social objects. The filtering uses predetermined thresholds tied to linking frequency and re-post or retweet activity, and the stored SVD-based representation is then used to perform queries.

Stated Advantages

Documented Applications