Lab

Experiments I'm running, tools anyone can use, and the outside work I build on — each labelled by its real stage and properly credited.

Experiments

ExperimentIn progress

Training and evaluating a Bengali model for claim detection and verification.

ExperimentIn progress

Transcription, claim extraction and source-checking for news video.

Established, openly documented work I use or recommend. It belongs to its authors; follow each link for terms of use.

Media Cloud
Media Ecosystems Analysis Group, UMass Amherst and Northeastern University · Open source (Apache-2.0); data under its terms of use
Open-source platform for searching and comparing online news coverage across outlets and countries.
The GDELT Project
Kalev H. Leetaru · Free for any use, with citation and a link back
Open database of world news events since 1979, updated every 15 minutes — for discovering reporting, not proving a claim.
Google Fact Check Tools API
Google · Google APIs Terms of Service (free API key)
Searches published fact-checks (ClaimReview) from fact-checking organisations worldwide.
FEVER: a Large-scale Dataset for Fact Extraction and VERification
Thorne, Vlachos, Christodoulopoulos and Mittal — NAACL 2018 · Wikipedia licence terms (CC BY-SA 3.0); code Apache-2.0
185,445 claims checked against Wikipedia — the standard benchmark for automated claim verification.
AVeriTeC: A Dataset for Real-world Claim Verification with Evidence from the Web
Schlichtkrull, Guo and Vlachos — NeurIPS 2023 · CC BY-NC 4.0 (non-commercial)
4,568 real-world claims from 50 fact-checking organisations, with evidence gathered from the web.

Whisper
OpenAI · MIT (code and model weights)
Open speech-recognition model for transcribing and translating audio — a starting point for news-video transcripts.

BanglaBERT: Language Model Pretraining and Benchmarks for Low-Resource Language Understanding Evaluation in Bangla
Bhattacharjee et al., CSE BUET NLP — Findings of NAACL 2022 · CC BY-NC-SA 4.0 (non-commercial research)
Bangla language-understanding model and benchmark suite; code and weights at github.com/csebuetnlp/banglabert.
BanglaNLG and BanglaT5: Benchmarks and Resources for Evaluating Low-Resource Natural Language Generation in Bangla
Bhattacharjee, Hasan, Ahmad and Shahriyar — Findings of EACL 2023 · CC BY-NC-SA 4.0 (non-commercial research)
Bangla text-generation model and benchmarks — useful for summarising and explaining claims in Bangla.