作者: Wendy Lehnert , Jonathan Aseltine , David Fisher , Stephen Soderland
DOI:
关键词: Crystal (programming language) 、 Natural language processing 、 Cover (telecommunications) 、 Information extraction 、 Machine-readable dictionary 、 Computer science 、 Artificial intelligence 、 Conceptual dictionary
摘要: One of the central knowledge sources an information extraction (IE) system IS a dictionary linguistic patterns that can be used to identify references relevant in text Automatic creation conceptual dictionaries is important for portability and scalability IE This paper describes CRYSTAL, which automatically induces "concept-node definitions" sufficient from training corpus Each these concept-node definitions generalized as far possible without producing errors, so minimum number entries cover positive instances Because it tests accuracy each proposed definition, CRYSTAL often surpass human intuitions creating reliable rules.