trec_eval ☆

« Back to VersTracker

Description:
Evaluation software used in the Text Retrieval Conference

Type: Formula | Tracked Since: Dec 28, 2025

Links: Homepage | formulae.brew.sh

Category: Developer tools

Tags: information-retrieval evaluation benchmarking search research

Install: brew install trec_eval

About:
TrecEval is the standard evaluation tool for information retrieval systems, developed for the Text Retrieval Conference (TREC). It calculates a wide range of performance metrics, such as precision and recall, by comparing system-generated rankings against human-judged relevance data. This tool is essential for benchmarking search algorithms and ensuring rigorous evaluation in research and production environments.

Key Features:

Computes standard IR metrics (Precision, Recall, MAP, NDCG)
Handles multi-level relevance judgments
Generates trec_eval output for detailed analysis
Supports multi-query evaluation

Use Cases:

Benchmarking search engine ranking algorithms
Academic research in information retrieval
Validating ML models for search relevance

Alternatives:

pytrec_eval – Python binding for trec_eval; faster for batch evaluation in Python pipelines
ranx – Modern Rust-based library with optimized metrics and Python bindings

Version History

Detected	Version	Rev	Change	Commit