Automatic Query Feedback using Related Words.
Stefan Rüger · 1999
. Our experiments for the ad hoc task of TREC 8 were centered around the question how to create an automatic query feedback from the documents returned by an initial query. 1 The Query Process 1.1 Preprocessing of the Documents In our retrieval experiments with N = 528; 155 articles, we folded all words to lowercase and indexed all words with a document frequency of no more than 30% and which were not one of 349 stop words. For each document i we compute a weighted document length of l i = s X j (t ij log(0:3N=d j )) 2 ; (1) where t ij is the term frequency of word j in document i and d j is the document frequency of word j. Additionally, we identied roughly 100 potentially interesting words per document which we stored along with the meta-data of the document at index time. For these words we only kept nouns and adjectives based on Brill's tagger (Brill 1994) with a medium document frequency: the noun had to appear in least three documents and in no more that 30% of all do...