---
id: trec-car-tools
version: "2.6"
license: BSD 3-Clause
license_treatment: permissive
maintenance: aging
---

# trec-car-tools — Support tools for TREC CAR participants. Also see trec-car.cs.unh.edu

License: permissive · Maintenance: aging · Downloads: 217.8K/mo

## What it is and what it does

trec-car-tools is a toolkit for participants in the TREC Complex Answer Retrieval track, a shared task in information retrieval research. It provides Python and Java libraries to deserialize Wikipedia data (articles, outlines, paragraphs) that have been encoded in CBOR binary format, and to serialize ranking results back into the standard TREC evaluation format. The package preserves Wikipedia-internal hyperlinks and metadata (categories, redirects, infoboxes, Wikidata IDs) during deserialization.

The typical workflow involves downloading TREC CAR data, using the package to parse it into Python or Java objects, running your retrieval or ranking algorithm, then using format_runs utilities to write results in the format expected by trec_eval and the official evaluation pipeline. It integrates with standard IR evaluation tools used in the research community.

Use it for:

- Parse Wikipedia paragraph collections and outlines for TREC CAR shared task participation
- Extract structured Wikipedia metadata (sections, hyperlinks, categories, infoboxes) for IR research
- Format and validate ranking results to submit to TREC CAR evaluation
- Integrate TREC CAR data into information retrieval pipelines supporting multiple format versions
- Analyze Wikipedia structure and entity-passage relationships for complex answer retrieval

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Provides Python and Java bindings to read TREC Complex Answer Retrieval track data in CBOR format, including paragraph collections, outlines, and Wikipedia articles with preserved hyperlinks.

Yes, if you are actively participating in the TREC CAR track or building research systems that depend on its data formats. The package has no runtime dependencies, installs easily, and carries no licensing restrictions. However, the aging maintenance status (last release 2022-02-01) means you should verify compatibility with your Python version and check whether the cbor library remains available before committing to a dependency.

## Install

pip install trec-car-tools
uv add trec-car-tools
poetry add trec-car-tools

## Installing trec-car-tools

Before you install: Low friction installation as a pure Python wheel with no runtime dependencies. However, the package is aging—last release was 2022-02-01, and the repository shows minimal recent activity.

License in practice: BSD 3-Clause is permissive and poses no restrictions on commercial or proprietary use, modification, or redistribution as long as the license and copyright notice are included.

Quickstart:

pip install trec-car-tools

import trec_car_tools

# Read paragraph collection or outline data in CBOR format
# See package documentation for specific module and function names

Requires the cbor library; install via 'conda install -c laura-dietz cbor=1.0.0' (Anaconda) or 'pip install cbor' (pip). Python 3.6 is the minimum supported version per classifiers.

Verify before relying:

- Whether the package works reliably with Python versions beyond 3.8 despite classifiers listing 3.6–3.8
- Current status and availability of the cbor dependency in modern Python environments
- Whether TREC CAR data formats v1.5, v2.0, and v2.6 remain in active use or if the track has evolved
- Specific module names and APIs available in the package for reading and writing data

## Package facts

- License: BSD 3-Clause (permissive)
- Python support: unspecified
- Install friction: low
- Maintenance: aging
- Downloads: 217.8K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

trec car data reader, cbor wikipedia parser, trec complex answer retrieval, paragraph collection tools, information retrieval dataset, trec eval integration, wikipedia outline processing, information-retrieval, shared-task, wikipedia-data

[View on SkillFed](https://skillfed.io/packages/trec-car-tools) · [View on PyPI](https://pypi.org/project/trec-car-tools/)