Improving the effectiveness of retrieval systems by information structures

摘要

A new information structure is presented that is aimed at increasing the effectiveness of an information retrieval system. This structure is a manually constructed knowledge representation providing additional information to the retrieval process. To analyze the properties of this new structure, the framework of first-order theories (model theory) is used. The results show that the construction and maintenance of such a structure are easier than those of a thesaurus. Furthermore, a retrieval function is presented that depends on such a new information structure. This retrieval function takes into account term dependencies (term associations) and it distinguishes between the different meanings of polysemic terms. This improves the effectiveness of the retrieval system since taking into account term dependencies increases the recall and interpreting polysemic indexing term increases the precision.