A Linguistic Approach to Information Retrieval

This report describes a series of experiments which were performed in Information Retrieval (IR) and especially in document retrieval. Our main goal was to improve the precision and recall of conventional document retrieval systems by means of Natural Language Processing (NLP). Although we chose the English language due to its weak morphology and simple syntax, the same techniques can also be used for other languages, but maybe with a lot of trouble for complex languages. The NLP that took place covered the morphological, lexical and syntactical part of English language, but not the semantical. Our system accepts English noun phrases as queries and extracts keywords from them. Techniques for morphological and lexical expansion of keywords were researched, implemented, applied and proved indispensable for high levels of recall. Conventional IR systems use keyword co-occurrence for retrieval of documents, and most of them do not examine the semantic relations between the keywords which ..