Relevance Reasoning in Text Retrieval

Relevance in a text retrieval task concerns the relevance of texts to a natural language query. The most difficult aspect of the task is to render the natural language of the text and the query into a form which permits computational relevance reasoning. English (or any language) must be rendered into logic before logical reasoning can be applied to it. Once this has been accomplished, relevance reasoning becomes a problem of determining the similarity of object and event descriptions. The basic similarity assignment in computational relevance judgment is sensitive to several complexities: granularity, ellipsis, domain, and genre.