Back to results
US 10,664,862 B1Granted
Topic inference based contextual content
AssigneeContextLogic Inc.
InventorsPiotr Szulczewski, Sheng Zhang, Jack Xie
Life of this patent
Estimating the term…
Abstract
Techniques to provide contextual content are disclosed. In various embodiments, a set of topics inferred programmatically from a corpus of raw content data is obtained. An input is received. A topic with which the input is associated is determined, based at least in part on the input and the set of topics. A contextual content associated with the topic is provided.
Claims30
A method, comprising:
- crawling, by one or more processors, through one or more electronic data sources to obtain a set of previously viewed online content data of a user, the one or more electronic data sources including one or more web pages or other online content viewed by the user;
- inferring, by the one or more processors, one or more topics associated with the set of previously viewed online content data, the inferring of the one or more topics comprising:
- extracting, by the one or more processors, metadata from the set of previously viewed online content data at least in part by parsing through source code associated with the one or more electronic data sources, the metadata is indicative of the one or more topics, and
- programmatically inferring the one or more topics by applying a natural language processing model to the set of previously viewed online content data, the inferring one or more topics includes weighing content obtained from a dynamic source of the one or more electronic data sources more heavily than content obtained from a static source of the one or more electronic data sources, and the inferring of the one or more topics is based at least in part on topics weighted based on the metadata;
- identifying, by the one or more processes, a hierarchical taxonomy of topics;
- generating, by the one or more processors, a set of remapped topics based at least in part on mapping, using a directed acyclic graph of the one or more inferred topics, each of the one or more inferred topics to a corresponding broader topic of the hierarchical taxonomy of topics, the set of remapped topics comprising a first remapped topic and a second remapped topic that is different than the first remapped topic;
- determining, by the one or more processors, that user input data from the user matches the first remapped topic;
- and
- providing, by the one or more processors, a first contextual content item on a web page being viewed by the user, the web page being associated with the second remapped topic, a second contextual content item associated with the second remapped topic being unavailable to be provided, the first contextual content item being associated with the first remapped topic, the first contextual content item being provided based at least in part on the second contextual content item of the second remapped topic generated from the one or more inferred topics being unavailable.
6 more elements
A system comprising:
- one or more processors of a machine;
- and
- a memory storing instructions that, when executed by the one or more processors, cause the machine to perform operations comprising:
- crawl through one or more electronic data sources to obtain a set of previously viewed online content data of a user, the one or more electronic data sources including one or more web pages or other online content viewed by the user;
- infer one or more topics associated with the set of previously viewed online content data, the inferring of the one or more topics comprising:
- extracting metadata from the set of previously viewed online content data at least in part by parsing through source code associated with the one or more electronic data sources, the metadata is indicative of the one or more topics, and
- programmatically inferring the one or more topics by applying a natural language processing model to the set of previously viewed online content data, the inferring one or more topics includes weighing content obtained from a dynamic source of the one or more electronic data sources more heavily than content obtained from a static source of the one or more electronic data sources, and the inferring of the one or more topics is based at least in part on topics weighted based on the metadata;
- identify a hierarchical taxonomy of topics;
- generate a set of remapped topics based at least in part on mapping, using a directed acyclic graph of the one or more inferred topics, each of the one or more inferred topics to a corresponding broader topic of the hierarchical taxonomy of topics, the set of remapped topics comprising a first remapped topic and a second remapped topic that is different than the first remapped topic;
- determine that user input data from the user matches the first remapped topic;
- and
- provide a first contextual content item on a web page being viewed by the user, the web page being associated with the second remapped topic, a second contextual content item associated with the second remapped topic being unavailable to be provided, the first contextual content item being associated with the first remapped topic, the first contextual content item being provided based at least in part on the second contextual content item of the second remapped topic generated from the one or more inferred topics being unavailable.
9 more elements
A non-transitory computer readable storage medium comprising instructions that, when executed by a machine, cause the machine to perform operations comprising:
- crawl through one or more electronic data sources to obtain a set of previously viewed online content data of a user, the one or more electronic data sources including one or more web pages or other online content viewed by the user;
- infer one or more topics associated with the set of previously viewed online content data, the inferring of the one or more topics comprising:
- extracting metadata from the set of previously viewed online content data at least in part by parsing through source code associated with the one or more electronic data sources, the metadata is indicative of the one or more topics, and
- programmatically inferring the one or more topics by applying a natural language processing model to the set of previously viewed online content data, the inferring one or more topics includes weighing content obtained from a dynamic source of the one or more electronic data sources more heavily than content obtained from a static source of the one or more electronic data sources, and the inferring of the one or more topics is based at least in part on topics weighted based on the metadata;
- identify a hierarchical taxonomy of topics;
- generate a set of remapped topics based at least in part on mapping, using a directed acyclic graph of the one or more inferred topics, each of the one or more inferred topics to a corresponding broader topic of the hierarchical taxonomy of topics, the set of remapped topics comprising a first remapped topic and a second remapped topic that is different than the first remapped topic;
- determine that user input data from the user matches the first remapped topic;
- and
- provide a first contextual content item on a web page being viewed by the user, the web page being associated with the second remapped topic, a second contextual content item associated with the second remapped topic being unavailable to be provided, the first contextual content item being associated with the first remapped topic, the first contextual content item being provided based at least in part on the second contextual content item of the second remapped topic generated from the one or more inferred topics being unavailable.
6 more elements
Dependencies are inferred from the claim text.