Context based resource relevance
AssigneeGoogle Inc.
InventorsAaron Shon, Matthew Zitzmann, Clyde D. McQueen, III, John Burge, Michael Skinner
Life of this patent
Estimating the term…
Abstract
A resource context system (“context system”) computes a contextual profile for a resource based on selection data for the resource and contextual profiles of other resources that are identified as relevant to the resource. The contextual profile includes values that specify measures of relevance of the resource to each of a plurality of corresponding topics. The contextual profile is provided to a processing system, such as an advertisement management system or a search system that can identify topics for which the resource is relevant based on the contextual profile.
Claims13
A computer-implemented method, comprising:
- identifying a reference resource and a contextual profile for the reference resource;
- determining that relevance feedback data for the reference resource has at least a threshold level of similarity to the contextual profile for the reference resource, the determination being based on a difference between the relevance feedback data and topics specified by the contextual profile, the relevance feedback data specifying one or more topics that users have identified as relevant to the reference resource;
- in response to the determination, computing, by a data processing device, a first contextual profile for a different resource based on the contextual profile for the reference resource;
- and
- providing the first contextual profile to a processing system that is configured to identify the different resource as relevant to at least one topic from the first contextual profile when at least one topic value in the first contextual profile meets a topic value threshold for the at least one topic, wherein:
- identifying the contextual profile for the reference resource comprises:
- identifying topic values for topics to which content of the reference resource has been identified as relevant by a clustering system;
- and
- identifying data specifying a cluster topic for a cluster to which a vector of terms for the resource is identified as relevant.
A computer-implemented method, comprising:
- identifying a reference resource and a contextual profile for the reference resource;
- determining that relevance feedback data for the reference resource has at least a threshold level of similarity to the contextual profile for the reference resource, the determination being based on a difference between the relevance feedback data and topics specified by the contextual profile, the relevance feedback data specifying one or more topics that users have identified as relevant to the reference resource;
- in response to the determination, computing, by a data processing device, a first contextual profile for a different resource based on the contextual profile for the reference resource;
- and
- providing the first contextual profile to a processing system that is configured to identify the different resource as relevant to at least one topic from the first contextual profile when at least one topic value in the first contextual profile meets a topic value threshold for the at least one topic, wherein:
- identifying a reference resource comprises identifying a resource that was sequentially selected for presentation relative to selection of the different resource;
- and
- computing a first contextual profile comprises:
- representing the reference resource and the different resource as nodes in a weighted graph, the node for the different resource being connected to the reference resource by a weighted edge representing sequential selections of the different resource and the reference resource;
- and
- computing the first contextual profile based on a function of the weighted edge connected to the node for the different resource and the reference contextual profile for the reference resource.
A computer-implemented method comprising:
- identifying a reference resource and a contextual profile for the reference resource;
- determining that relevance feedback data for the reference resource has at least a threshold level of similarity to the contextual profile for the reference resource, the determination being based on a difference between the relevance feedback data and topics specified by the contextual profile, the relevance feedback data specifying one or more topics that users have identified as relevant to the reference resource;
- in response to the determination, computing, by a data processing device, a first contextual profile for a different resource based on the contextual profile for the reference resource;
- normalizing topic values of the first contextual profile to normalized topic values for the first contextual profile, the normalized topic values being values on a normalized scale;
- adjusting, to a baseline value, each of the normalized topic values that is less than a first topic value threshold;
- re-normalizing the normalized topic values having values that are different than the baseline value subsequent to the adjusting;
- and
- re-computing the first contextual profile based on a function of the re-normalized topic values and the contextual profile for the reference resource.
A system, comprising:
- a data store storing contextual profiles for resources, the contextual profiles specifying topic values that represent measures of relevance of the resources to each of a plurality of topics and
- a resource context system coupled to the data store, the resource context system including at least one processor configured to perform operations comprising:
- identifying, from the data store, a reference resource and a contextual profile for the reference resource;
- determining that relevance feedback data for the reference resource has at least a threshold level of similarity to the contextual profile for the reference resource, the determination being based on a difference between the relevance feedback data and topics specified by the contextual profile, the relevance feedback data specifying one or more topics that users have identified as relevant to the reference resource;
- in response to the determination, computing a first contextual profile for a different resource based on the contextual profile for the reference resource;
- and
- providing the first contextual profile to a processing system that is configured to identify the different resource as relevant to at least one topic from the first contextual profile when at least one topic value in the first contextual profile meets a topic value threshold for the at least one topic, wherein computing the first contextual profile comprises:
- representing the reference resource and the different resource as nodes in a weighted graph, the node for the different resource being connected to the reference resource by a weighted edge representing sequential selections of the different resource and the reference resource;
- and
- computing the first contextual profile based on a function of the weighted edge and the contextual profile for the reference resource.
A system, comprising:
- a data store storing contextual profiles for resources, the contextual profiles specifying topic values that represent measures of relevance of the resources to each of a plurality of topics and
- a resource context system coupled to the data store, the resource context system including at least one processor configured to perform operations comprising:
- identifying, from the data store, a reference resource and a contextual profile for the reference resource;
- determining that relevance feedback data for the reference resource has at least a threshold level of similarity to the contextual profile for the reference resource, the determination being based on a difference between the relevance feedback data and topics specified by the contextual profile, the relevance feedback data specifying one or more topics that users have identified as relevant to the reference resource;
- in response to the determination, computing a first contextual profile for a different resource based on the contextual profile for the reference resource;
- normalizing topic values of the first contextual profile to normalized topic values for the first contextual profile, the normalized topic values being values on a normalized scale;
- adjusting, to a baseline value, each of the normalized topic values that is less than a first topic value threshold;
- re-normalizing the topic values based on the adjusted topic values and
- providing the first contextual profile to a processing system that is configured to identify the different resource as relevant to at least one topic from the first contextual profile when at least one topic value in the first contextual profile meets a topic value threshold for the at least one topic.
A non-transitory computer readable storage medium encoded with a computer program comprising instructions that when executed operate to cause a computer to perform operations:
- identifying a reference resource and a contextual profile for the reference resource;
- determining that relevance feedback data for the reference resource has at least a threshold level of similarity to the contextual profile for the reference resource, the determination being based on a difference between the relevance feedback data and topics specified by the contextual profile, the relevance feedback data specifying one or more topics that users have identified as relevant to the reference resource;
- in response to the determination, computing, by a data processing device, a first contextual profile for a different resource based on the contextual profile for the reference resource;
- normalizing the topic values of the first contextual profile to normalized topic values for the first contextual profile, the normalized topic values being values on a normalized scale;
- adjusting, to a baseline value, each of the normalized topic values that is less than a first topic value threshold;
- and
- re-normalizing the normalized topic values having values that are different than the baseline value subsequent to the adjusting;
- providing the first contextual profile to a processing system that is configured to identify the different resource as relevant to at least one topic from the first contextual profile when at least one topic value in the first contextual profile meets a topic value threshold for the at least one topic.
Dependencies are inferred from the claim text.