The Mirror DBMS at TREC-9

The Mirror DBMS is a prototype database system especially designed for multimedia and web retrieval. From a database perspective, this year's purpose has been to check whether we can get sufficient efficiency on the larger data set used in TREC-9. From an IR perspective, the experiments are limited to rather primitive web-retrieval, teaching us that web-retrieval is (un?-)fortunately not just retrieving text from a different data source. We report on some limited (and disappointing) experiments in an attempt to benefit from the manually assigned data in the metatags. We further discuss observations with respect to the effectiveness of title-only topics.

• Type: Conference
• Year: 2000
• Where: TREC
• Authors: Arjen P. de Vries
• Keywords: Mirror DBMS, Prototype Database, Sufficient Efficiency, TREC 2000, TREC 2008