Bi-directional automata to extract complex phrases from texts

摘要

This paper presents an experiment to develop natural-language tools to improve the quality of documents. These softwares are using finite-state automata enriched with notions of proximity, option-ality and contextual information. They are called bi-directional because they need to parse a sequence not only from the left to right-hand side of a sentence, but on both sides of a word. This method improves efficiency.