Average gain ratio which is based on the idea of cumulative gain and is used for evaluation with multiple relevance levels, is proposed. Cumulative gain assumes that an imaginary user scans the ranked output from the top, and 'gains' an additional score each time he finds a relevant document. Average gain ratio was proposed as a modification of weighted average precision. The results show that average gain ratio is more suitable for averaging across topics and is more robust to the variation in the number of relevant documents at each relevance level.
|ジャーナル||SIGIR Forum (ACM Special Interest Group on Information Retrieval)|
|出版ステータス||Published - 2003 12月 1|
|イベント||Proceedings of the Twenty-Sixth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2003 - Toronto, Ont., Canada|
継続期間: 2003 7月 28 → 2003 8月 1
ASJC Scopus subject areas