Classification of documents
AssigneeFirstrain, Inc.
InventorsAshutosh Joshi, Martin Betz, Rajiv Arora, Rakesh Kumar Srivastava, David Cooke
Life of this patent
Estimating the term…
Abstract
Some embodiments provide a method for evaluating a content segment for relevancy to several of categories. The method retrieves the content segment. For each of the several categories, the method determines the relevancy of the content segment to the category by using a scoring model for the category. The scoring model accounts for (i) the presence of key word sets in the content segment and (ii) the context of the key word sets in the content segment. For each of the several categories, the method tags the content segment when the content segment is determined as relevant to the category.
Claims13
A method for evaluating a content segment for relevancy to a plurality of categories, the method comprising:
- retrieving the content segment;
- for each of a plurality of categories, calculating a score for the content segment to determine the relevancy of the content segment to the category by using a context-based model for the category, said context-based model comprising (i) a set of groups of word sets, each group of word sets comprising a key word set and a second word set, (ii) scores for the groups of word sets, and (iii) a definition of context that specifies when a second word set is in a context of a key word set, wherein the context for a particular key word set is based on a specified relationship, within the content segment, between the particular key word set and a second word set, wherein the calculating comprises:
- identifying each key word set from the context-based model that is in the content segment;
- for each identified key word set, associating the key word set and each word set within the context of the key word set in the content segment as a different group of word sets;
- and
- aggregating scores for each of the associated groups of word sets to calculate the score for the content segment;
- and
- for each of the plurality of categories, tagging the content segment with a category tag to signify that the content segment is relevant to the category when the content segment is determined to be relevant to the category.
A system for evaluating a plurality of content segments for relevancy to a plurality of different categories, the system comprising:
- a set of processing units;
- and
- a machine readable medium storing:
- a crawler which when executed by at least one of the processing units searches a network for new content segments on a regular basis and downloads the new content segments;
- a content segment evaluator which when executed by at least one of the processing units determines whether the new content segments are relevant to each of the plurality of different categories by using content relevance models for the plurality of different categories, wherein at least one of the content relevance models for a particular category is a context-based model comprising (i) a set of groups of word sets, each group of word sets comprising a key word set and a second word set, (ii) scores for the groups of word sets, and (iii) a definition of context that specifies when a second word set is in the context of a key word set, wherein the context-based model determines whether a new content segment is relevant by:
- identifying each key word set from the context-based model that is in the new content segment;
- for each identified key word set, associating the key word set and each word set within the context of the key word set in the content segment as a different group of word sets;
- and
- aggregating scores for each of the associated groups of word sets to calculate a relevancy score for the new content segment;
- and
- a content segment tagger which when executed by at least one of the processing units tags a particular content segment with a particular category and the relevancy score for the category when the content segment is relevant to the category.
Dependencies are inferred from the claim text.