Hypothesis testing for the risk-sensitive evaluation of retrieval systems
- Submitting institution
-
University of Glasgow
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 11-02723
- Type
- E - Conference contribution
- DOI
-
10.1145/2600428.2609625
- Title of conference / published proceedings
- 37th International ACM SIGIR Conference on Research and Development in Information Retrieval
- First page
- 23
- Volume
- -
- Issue
- -
- ISSN
- -
- Open access status
- -
- Month of publication
- -
- Year of publication
- 2014
- URL
-
http://eprints.gla.ac.uk/95215/
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
2
- Research group(s)
-
-
- Citation count
- 15
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- ORIGINALITY: Uses novel inferential analysis techniques to ensure that a deployed search engine ranking strategy is robust across all queries, i.e.. it increases average effectiveness while ensuring that no single query falls below a benchmark performance level. Rigor: Experiments and analysis are conducted using a NIST standard dataset, and a very large learning-to-rank dataset from Microsoft Research. SIGNIFICANCE: Published at top IR conference. Enhancing robustness of learning-to-rank approaches widely used in commercial web search engines (e.g. Bing) ensures better end-user satisfaction. This paper's outcome was integrated into the widely-used/cited Jforests open-source learning-to-rank library (https://github.com/yasserg/jforests).
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -