Back to results
US 2008/0168019 A1Published application
Document clustering based on entity association rules
AssigneeMicrosoft Corporation
InventorsWisam Dakka, Silviu-Petru Cucerzan
Life of this patent
Estimating the term…
Abstract
For each document in a document set, entities are identified and a set of association rules, based on appearance of the entities in the paragraphs of the documents in the set, are derived. Documents are clustered based on the association rules. As documents are added to the clusters, additional association rules specific to the clusters can optionally be derived as well.
Claims20
A method of generating a document clustering system, comprising:
- identifying entities in each document in a set of documents;
- obtaining association rules, corresponding to an initial set of documents, based on co-occurrences of the entities in the initial set of documents, each of the association rules identifying one or more entities that co-occur with other entities in the initial set of documents with a frequency that meets a threshold;
- and
- assigning the documents to a first set of clusters, based on the association rules derived from the initial set of documents.
1 more element
A method of grouping documents by topic, comprising:
- matching occurrences of entities in a document against sets of association rules, each set of association rules corresponding to one of a first set of groups, and each association rule indicating that an entity occurs in a document if one or more other entities appear in the document, with a desired frequency;
- and
- grouping the documents into a given group in the first set of groups based on a frequency with which the occurrences of entities in the document match a set of association rules corresponding to the given group.
A document clustering system, comprising:
- a clustering component configured to cluster a received document into one of a set of clusters, each cluster having a corresponding set of association rules, each association rule indicating an entity that occurs in a document given an occurrence of one or more other entities in the document, the clustering component matching the association rules against occurrences of entities in the received documents and selecting a cluster within which to place the received document based on matching association rules;
- an entity identifier configured to identify the occurrences of entities in individual paragraphs of the received document;
- and
- an association rules generator configured to generate additional association rules for the selected cluster based on inclusion of the received document in the selected cluster.
1 more element
Dependencies are inferred from the claim text.