On Intra-page and Inter-page Semantic Analysis of Web Pages
Jun Wang | Jicheng Wang | Gangshan Wu | Hiroshi Tsuda
Proceedings of the 17th Pacific Asia Conference on Language, Information and Computation
In the domain of artificial intelligence, the pattern of information flow varies drastically from one context to another. To capture this diversity of information flow, a natural-language processing (NLP) system should consist of modules of constraints and one general constraint solver to process all of them; there should be no specialized procedure module such as a parser and a generator. This paper presents how to implement such a constraint-based approach to NLP. Dependency Propagation (DP) is a constraint solver which transforms the program (=constraint) represented in terms of logic programs. Constraint Unification (CU) is a unification method incorporating DP. cu-Prolog is an extended Prolog which employs CU instead of the standard unification. cu-Prolog can treat some lexical and grammatical knowledge as constraints on the structure of grammatical categories, enabling a very straightforward implementation of a parser using constraint-based grammars. By extending DP, one can deal efficiently with phrase structures in terms of constraints. Computation on category structures and phrase structures are naturally integrated in an extended DP. The computation strategies to do all this are totally attributed to a very abstract, task-independent principle: prefer computation using denser information. Efficient parsing is hence possible without any parser.