BEIR CQADupStack — Android (orgrctera/beir_cqadupstack_android)

Overview

This release packages the Android subforum slice of CQADupStack from the BEIR (Benchmarking IR) benchmark as a table-oriented dataset for retrieval evaluation and tooling (e.g. Langfuse-exported runs).

CQADupStack is a community question answering (cQA) resource built from Stack Exchange forums: posts are paired with duplicate-question annotations so systems can be evaluated on finding earlier questions that address the same information need. The Android subset restricts the corpus and queries to the android.stackexchange.com forum—typical topics include device setup, apps, rooting, OEM behavior, and connectivity.

In the BEIR formulation, each test query is matched against a corpus of forum posts (questions with titles and body text, plus metadata such as tags in the upstream BEIR distribution). Relevance judgments (qrels) mark which corpus documents are true duplicates (or duplicate-cluster members used as positives) for that query. Evaluators rank retrieved documents and measure overlap with these IDs using standard IR metrics.

Lineage: CQADupStack (original benchmark) → BEIR cqadupstack/android → this Hub dataset (orgrctera/beir_cqadupstack_android).

Scale (BEIR / ir-datasets): the Android configuration includes on the order of ~699 test queries and ~23K corpus documents (full corpus is indexed at evaluation time; this table stores query rows with gold document IDs, not the corpus text).

Task

• Task type: Retrieval — duplicate-question retrieval on the CQADupStack Android collection (BEIR sub-benchmark cqadupstack_android).
• Input (input): Natural-language query text (a Stack Exchange question title/body-style string as distributed in BEIR).
• Reference (expected_output): A JSON string encoding the list of relevant corpus document IDs with binary relevance scores (typically 1 for relevant pairs), e.g. [{"id": "1120", "score": 1}, ...].
• Metadata: metadata.query_id is the BEIR query identifier; metadata.split is test for this release.

The retrieval system’s job is to return the correct corpus document IDs for each query when scored against the full CQADupStack Android corpus distributed with BEIR (not duplicated row-wise in this table).

Background

CQADupStack (original dataset)

Doris Hoogeveen, Karin M. Verspoor, and Timothy Baldwin introduced CQADupStack as a benchmark for community question answering, with emphasis on duplicate question detection and related retrieval/classification setups. The resource spans twelve Stack Exchange subforums; each subforum provides annotated duplicate relations derived from site curation (e.g. “linked” / duplicate closures), enabling comparable train/test splits for retrieval and classification experiments. The underlying posts come from Stack Exchange data dumps (the original release notes reference a September 26, 2014 dump).

Community question answering (cQA) forums accumulate large volumes of user-generated knowledge… Duplicate question detection—identifying whether a new question has already been answered—is central to maintaining these archives.
(Paraphrased theme; see the ADCS 2015 paper below for the full problem statement and methodology.)

• Paper: CQADupStack: A Benchmark Data Set for Community Question-Answering Research (ADCS 2015) — PDF mirror: University of Melbourne
• Website: nlp.cis.unimelb.edu.au/resources/cqadupstack
• Repository: D1Doris/CQADupStack

BEIR reformulation

BEIR (Thakur et al., NeurIPS 2021 Datasets & Benchmarks) repackages CQADupStack (per subforum, including android) into a unified layout: corpus (JSONL: _id, title, text, …), queries (JSONL: _id, text, …), and qrels (TSV). That format supports zero-shot comparison of lexical, sparse, dense, and reranking retrievers across heterogeneous tasks.

This release

Rows were exported from Langfuse (CTERA AI evaluation pipeline) in a flat, Parquet-friendly schema: one row per query with gold relevant document IDs in expected_output for downstream scoring and observability.

Data fields

Column	Type	Description
id	string	Stable UUID for this row in this Hub release.
input	string	Query text (natural-language question).
expected_output	string	JSON string: list of objects {"id": "<corpus-doc-id>", "score": <int>} — qrels for that query.
metadata.query_id	string	BEIR CQADupStack Android query identifier.
metadata.split	string	Split name: test.

Splits

Split	Rows
test	699
Total	699

Examples

Real rows from this dataset (IDs and text as published on the Hub).

Example 1 — multi-duplicate query

• input: Symbolic link to Dropbox
• metadata.query_id: 47290
• metadata.split: test
• expected_output:

[
  {"id": "1120", "score": 1},
  {"id": "31804", "score": 1},
  {"id": "14873", "score": 1},
  {"id": "38645", "score": 1}
]

Example 2 — single relevant document

• input: Does anyone else have issues with Google Talk for Android always "losing connection to server"?
• metadata.query_id: 10118
• metadata.split: test
• expected_output:

[
  {"id": "4345", "score": 1}
]

References and citations

BEIR benchmark (aggregation & protocol)

Nandan Thakur, Nils Reimers, Andreas Rücklé, Abhishek Srivastava, Iryna Gurevych. BEIR: A Heterogeneous Benchmark for Zero-shot Evaluation of Information Retrieval Models. NeurIPS 2021 Datasets and Benchmarks Track.

Abstract (excerpt): We introduce Benchmarking-IR (BEIR), a robust and heterogeneous evaluation benchmark for information retrieval. We leverage a careful selection of 18 publicly available datasets from diverse text retrieval tasks and domains and evaluate 10 state-of-the-art retrieval systems including lexical, sparse, dense, late-interaction and re-ranking architectures on the BEIR benchmark…

• Paper: arXiv:2104.08663 · OpenReview
• Code / data: beir-cellar/beir

@inproceedings{thakur2021beir,
  title={{BEIR}: A Heterogeneous Benchmark for Zero-shot Evaluation of Information Retrieval Models},
  author={Thakur, Nandan and Reimers, Nils and R{\"u}ckl{\'e}, Andreas and Srivastava, Abhishek and Gurevych, Iryna},
  booktitle={Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2)},
  year={2021},
  url={https://openreview.net/forum?id=wCu6T5xFjeJ}
}

CQADupStack (original Android / multi-forum resource)

Doris Hoogeveen, Karin M. Verspoor, Timothy Baldwin. CQADupStack: A Benchmark Data Set for Community Question-Answering Research. Proceedings of the 20th Australasian Document Computing Symposium (ADCS 2015).

• DOI: 10.1145/2838931.2838934
• PDF: adcs2015.pdf

@inproceedings{hoogeveen2015cqadupstack,
  author    = {Hoogeveen, Doris and Verspoor, Karin M. and Baldwin, Timothy},
  title     = {{CQADupStack}: A Benchmark Data Set for Community Question-Answering Research},
  booktitle = {Proceedings of the 20th Australasian Document Computing Symposium},
  year      = {2015},
  pages     = {3:1--3:8},
  publisher = {ACM},
  doi       = {10.1145/2838931.2838934}
}

Related tooling

• ir-datasets catalog entry: beir/cqadupstack/android (query/doc counts and export helpers).

---

Dataset card maintained for the orgrctera/beir_cqadupstack_android Hub repository.