medlars

Sinopsis

MEDLARS: A small data base, monolingual, for teaching pourposes only.

Description

MEDLARS is a monolingual corpus, with 1033 documents, 30 queries and human relevance judgements. The queries length varies between 2 to 62 words, with an average of 20. The number of relevant documents per query varies between 9 to 39 with an average of 23.

If you look at the data directory we now have the following qrel files for the med set:
qrels.text - original
med.training.smart - first 500 articles
med.test.smart - second 533 articles
med1.test.smart - second 533 articles, renumbered to start at DID 1
queries.text - 30 queries
qrels.text - all qrels
qrels.text.2nd_half - qrels for did's > 500
qrels.text.1st_half - qrels for did's 1-500
qrels.text.2nd_half_from1 - qrels for did's > 500, renumbered to 1-533

Links

oslo: /usr9/ir/tomp/ir11741/corpus/med/