作者: Fergus Kelledy , Ruairi O'Donnell , Alan F. Smeaton
DOI:
关键词:
摘要: In this paper we describe work done as part of the TREC-4 benchmarking exercice by a team from Dublin City University. had 3 activities follows : on improving efficiency standard SMART like query processing have applied various thresholding processes to postings list an inverted file and limited number document score accumulators available during processing. The first run submitted for evaluation in used our best set acumulator parameters ; second is based upon expansion using terms WordNet. Essentially, each original term determine its level specificity or abstraction broad add more specific terms, broader ones in-between both narrower terms. When expanded then delete all order judged pool, documents that would find nat been found other retrieval. This DCU952 third was Spanish data. We ran entire corpus through POS tagger indexed (and query) combination base form non stopwords plus their class. Retrieval performed with extra weights depending