IR and AI: traditions of representation and anti-representation in information processing
Yorick Alexander Wilks · 1999
The paper examines and contrasts three traditions of representation for information access tasks: (i) a statistical, distributional (and non-representational) approach, the core of information retrieval (IR) but increasingly used in recent natural language processing (NLP) applications; (ii) a linguistic, rule-driven, approach, which as been the core of most information extraction (IE) work; and (iii) artificial intelligence (AI) taken more narrowly (in a broad form it covers (i) and (ii) above) to mean knowledge-based processes. (4 pages)