Information Retrieval (IR) System Evaluation Task
Jump to navigation
Jump to search
An Information Retrieval (IR) System Evaluation Task is a system evaluation task that assesses the effectiveness of an Information Retrieval System.
- Context:
- It can (typically) involve the measurement of Relevance of the retrieved documents to a given Search Query.
- It can (typically) include IR Evaluation Measures such as Precision, Recall, F-Measure, and Mean Average Precision (MAP).
- It can (often) involve Benchmarking against standard datasets like TREC Collections.
- It can include Efficiency Measures, such as response time and computational resource usage.
- It can be conducted in Test Collections with predefined sets of documents, queries, and relevance judgments.
- It can involve User Study-based evaluations to measure User Satisfaction and Usability.
- It can be influenced by Information Retrieval Models, Query Processing techniques, and Indexing Strategy.
- It can involve A/B Testing to compare different retrieval algorithms or system configurations.
- …
- Example(s):
- an evaluation of a Search Engine's (e.g. using the TREC ad hoc task dataset).
- an evaluation of a Legal Document Retrieval System.
- an evaluation of a Corporate Document Retrieval System.
- …
- Counter-Example(s):
- An Q/A System Evaluation Task, which measures computational efficiency without considering retrieval effectiveness.
- A Database Query Performance Evaluation, which focuses on the speed and correctness of query execution but not on the relevance of the results.
- …
- See: Precision, Recall, F-Measure, Mean Average Precision, Information Retrieval System, Relevance, Test Collection, Benchmarking, TREC, User Study, Information Retrieval Model, A/B Testing.
References
2022
- (Thomas et al., 2022) ⇒ Paul Thomas, Gabriella Kazai, Ryen White, and Nick Craswell. (2022). “The Crowd is Made of People: Observations from Large-scale Crowd Labelling.” In: Proceedings of the 2022 Conference on Human Information Interaction and Retrieval. doi:10.1145/3498366.3505815