Given their simple nature, the success of keyword based retrieval systems is astonishing. Although these methods seemingly only process words (and their word counts), they rely on and credit most of their success to the keyword’s implicit semantics. In fact each keyword is a representation of a thought, or concept. Recent developments in NLP make it possible to use bigger syntactical units (like noun phrases) for IR purposes. The obtained phrases (and their subphrases) can be structured into a lithoid which forms an ideal starting point for feedback mechanisms like ‘Query By Navigation’ [1]. This paper will go even further and tries to involve the phrases’ support into the structure using a mathematical theory called ‘formal concept analysis’ [2]. The resulting concept lattice shows great similarity to the original lithoid and, since it consists of formal concepts, it leads the way to handle phrase semantics.
F. Grootjen