Understanding and Predicting Characteristics of Test Collections in Information Retrieval

TL;DR

This paper analyzes how the number of participants and other factors influence the quality of test collections in information retrieval evaluations, and demonstrates that the reusability of these collections can be accurately predicted over successive years.

Contribution

It provides empirical insights into factors affecting test collection quality and introduces methods to predict reusability based on historical data.

Findings

Number of teams significantly impacts test collection quality

Reusability of test collections can be predicted with high accuracy

Factors like participant diversity influence evaluation outcomes

Abstract

Research community evaluations in information retrieval, such as NIST's Text REtrieval Conference (TREC), build reusable test collections by pooling document rankings submitted by many teams. Naturally, the quality of the resulting test collection thus greatly depends on the number of participating teams and the quality of their submitted runs. In this work, we investigate: i) how the number of participants, coupled with other factors, affects the quality of a test collection; and ii) whether the quality of a test collection can be inferred prior to collecting relevance judgments from human assessors. Experiments conducted on six TREC collections illustrate how the number of teams interacts with various other factors to influence the resulting quality of test collections. We also show that the reusability of a test collection can be predicted with high accuracy when the same document collection is used for successive years in an evaluation campaign, as is common in TREC.