Full text retrieval based on syntactic similarities

摘要

Natural language queries are a basic requirement for modern information retrieval systems. We can therefore base retrieval in text information systems on text comparison using syntactic traces. The syntactic trace of a text is the set of all overlapping n-grams of this text. Thus, retrieval is done by comparing the n-gram sets. We define here a syntactic similarity function consisting of a direct and an indirect factor. That the presented theory is useful for the retrieval of information in natural language information systems, is shown by the results of the prototype TRIGIR based on trigrams.