Method and system for identifying user issues in forum posts based on discourse analysis

A system for categorizing issues comprising:

• a domain term vocabulary which includes a plurality of domain terms;
• memory which, for each of a plurality of categories of issue, stores at least one discourse pattern for identify patterns that are met by a text sequence, at least one of the patterns specifying that an instance of one of the domain terms be present in the text sequence for the pattern to be met;
• a syntactic parser which identifies syntactic dependencies in an issue comprising a text sequence;
• a discourse pattern extraction component which determines which of the discourse patterns are met by the text sequence;
• a categorization component which categorizes the text sequence based on the discourse patterns which are met;
• and
• an output component which outputs information based on the categorization of the text sequence.

A method for categorizing an issue comprising:

• providing a vocabulary of domain terms;
• for each of a plurality of categories of issue, providing at least one discourse pattern for identify text sequences that met the pattern, the issue categories including at least one information request category and an anomaly category, at least one of the anomaly category discourse patterns being met when an instance of one of the domain terms in the domain term vocabulary is in a syntactic relation with another term in the in the text sequence;
• receiving a text sequence;
• with a processor, categorizing the text sequence based on the discourse patterns that are met by the text sequence;
• and
• where the text sequence meets a discourse pattern for the anomaly category, processing the text sequence with a first procedure and where the text sequence meets a discourse pattern for the information request category, processing the text sequence with a second procedure, different from the first procedure.