Effective and Efficient Retrieval from Large and Dynamic Document Collections.

fficiency is conflicting with the retrieval effectiveness and with the retrieval efficiency. In the case of static data collections, the retrieval effectiveness criterion is usually met by means of weighted retrieval including relevance feedback [4], [9] whereas the retrieval efficiency criterion is met by precomputing the document descriptions and by organizing these descriptions as an inverted file [5]. In the case of dynamic data collections, however, a transaction updating the inverted file may block other retrieval and update transactions for an unacceptable long time because adding the description of a new document to an inverted file is time consuming, particularly if the document is long. A possible remedy is to use signatures instead of an inverted file [2]. Signatures can be updated efficiently; however, they do not allow document feature weighting and hence, the retrieval effectiveness of the signature based method is inferior to a fully weighted retrieval me