Using core words to extract key phrases from documents

Abstract

Described is a technology, such as for use in information retrieval, by which key phrases (e.g., the phrases most relevant to a document) are extracted from a document based upon core words in that document (e.g., the words most relevant to the document). Various relevance features of each candidate word may be used to score and rank the candidate words relative to one another and thereby determine the core word or core words. The core word or words may be used to filter a document's phrases into candidate phrases, and/or used to determine core word feature values associated with each candidate phrase. The features of each candidate phrase, one or more of which may be based on the presence or absence of core words in the candidate phrase, are used to rank the candidate phrases, with the top-ranked candidate phrases being the key phrases associated with the document.