TREC 7 Ad Hoc, Speech, and Interactive tracks at MDS/CSIRO.

Abstract

at the conference, jointly participated with newcomers CSIRO. Together we completed runs in three tracks: adhoc, interactive, and speech. 2 Ad-hoc task In TREC-5 we used document retrieval based on arbitrary passages [8, 9], or xed-length passages that could start at any word position. Although far from the best runs in TREC-5, these results were promising, in particular for long documents. In TREC-6 we continued with arbitrary passages, but our main emphasis was on comprehensive factor analysis of successful automatic query expansion and re nements methods in the context of the vector space model [5]. This year we have re ned the MG retrieval system to include Rocchio-based relevance feedback. Also, phrase matching has been added. We have continued to use arbitrary passages and combination of evidence for document retrieval. 2.1 System description An in-house version of the MG retrieval system has been used for all experiments. All experiments were carried out on an Intel Pentium II (300 Mhz) with a single processor and 256 Mb of physical memory. Queries and documents were matched using the Okapi formulation [13]: with wd;t: and wq;t: sim(q; d) =X k1 [(1, b)+b (k3 +1) fq;t k3 + fq;t t2q^d (k1 +1) fd;t log wd;t wq;t (1)