Effects of Inconsistent Relevance Judgments on Information Retrieval Test Results: A Historical Perspective
Open accessThe DOI matches its Crossref record; 1 of the 1 references it cites were found in a registry or an index.
Abstract
The main objective of information retrieval (IR) systems is to retrieve information or information objects relevant to user requests and possible needs. In IR tests, retrieval effectiveness is established by comparing IR systems retrievals (systems relevance) with users’ or user surrogates’ assessments (user relevance), where user relevance is treated as the gold standard for performance evaluation. Relevance is a human notion, and establishing relevance by humans is fraught with a number of problems—inconsistency in judgment being one of them. The aim of this critical review is to explore the relationship between relevance on the one hand and testing of IR systems and procedures on the other. Critics of IR tests raised the issue of validity of the IR tests because they were based on relevance judgments that are inconsistent. This review traces and synthesizes experimental studies dealing with (1) inconsistency of relevance judgments by people, (2) effects of such inconsistency on results of IR tests and (3) reasons for retrieval failures. A historical context for these studies and for IR testing is provided including an assessment of Lancaster’s (1969) evaluation of MEDLARS and its unique place in the history of IR evaluation.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
The CRANFIELD TESTS ON INDEX LANGUAGE DEVICES
CLEVERDON, C. (1967) · Aslib Proceedings · doi:10.1108/eb050097 · Cited by 509 · cited 1×
“Thirty-three types of index languages were investigated starting with single terms and then adding word forms and synonyms; broader, related, and narrower terms; and term phrases, hierarchies, and combinations thereof, with alterations of levels of specificity and exhaustivity of indexing (Cleverdon, 1967).”
Background
Cited by
Large Language Models can Accurately Predict Searcher Preferences
Thomas, P., Spielman, S., Craswell, N., & Mitra, B. (2024) · Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval · doi:10.1145/3626772.3657707 · Cited by 168
A framework for evaluating automatic indexing or classification in the context of retrieval
Golub, K., Soergel, D., Buchanan, G., Tudhope, D., Lykke, M., & Hiom, D. (2016) · Journal of the Association for Information Science and Technology · doi:10.1002/asi.23600 · Cited by 49
Hjørland, B. (2015) · Journal of the Association for Information Science and Technology · doi:10.1002/asi.23250 · Cited by 47
“Regrettably, failure tests are no longer conducted, mostly because they are complex, very time consuming, and CANNOT be done by a computer” (Saracevic, 2008, p. 772).”
BackgroundBenchmarking LLM-based Relevance Judgment Methods
Arabzadeh, N., & Clarke, C. L. A. (2025) · Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval · doi:10.1145/3726302.3730305 · Cited by 46
Metadata
- Authors
- Saracevic, T.
- Venue
- Library Trends
- Publisher
- Project MUSE
- Published
- 2008-03
- Volume
- 56
- Issue
- 4
- Pages
- 763-783
- ISSN
- 0024-2594 (ISSN-L)
- 1559-0682
- OpenAlex
- W2102500012
- Cited by
- 44
- References
- 1
- FWCI
- 10.36
- Type
- journal article
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.