The Chronological Information Extraction System (CHESS) performs two main functions to provide the user with rich contextual time-stamped information about relevant people, organizations, countries, or any other named entities. First, it extracts information about named entities and compiles a chronological outline. Second, it acts as a question answering system, allowing the user to find information of interest. In this capacity, the knowledge automatically extracted by CHESS can be examined in different information access modes: (1) as a browser, for information seekers who do not have specific queries; (2) as a question-answering system, for users with specific queries. Information is extracted from free text utilizing natural language processing techniques. CHESS takes advantage of the common practice of writers placing information-rich linguistic constructions in close proximity to named entities. First, the proper name is recognized, and then the surrounding information is used to develop relationships to other names entities. In this manner, concept-relation-concept (CRC) triples are formed and stored in an object database. Complex histories of events in relation to a given named entity are constructed from these CRC triples, utilizing techniques from conceptual graph theory.
展开▼
