The paper presents an approach to information retrieval for a specific domain of patent claims. The approach rests on utilizing predicate-argument structure representation of full-text documents for organizing a fine-tuned retrieval of semantic information. Indexing and search rely on a versatile data from a thoroughly elaborated linguistic knowledge base. The procedure of automatic indexing resulting in forming the predicate-argument structure of a patent claim is presented. The predicate-argument structure is conceived as a structural unit of the indexed document. The search strategy based on mapping predicate-arguments structures of the documents and query representations is described.
The Web Journal of Formal, Cognitive and Computational Linguistics. - Issue 8. - 2006