Fact Checking in Low Resource Languages: Creating Translation-Based Datasets and Benchmarking for Farsi
Bibliographic record
Abstract
Misinformation poses significant risks to public health, trust, and social cohesion. However, fact-checking datasets and models remain scarce in low-resource languages (LRLs) such as Amharic, Burmese, Chechen, Farsi (Persian), Maguindanao, Pashto, and Uighur. Creating large, manually curated datasets in these languages is often prohibitively expensive. This paper proposes a resource-efficient method for building fact-checking datasets in LRLs using machine translation. Specifically, we introduce FNC-1F, a Farsi dataset created by translating the widely used Fake News Challenge dataset (FNC-1) from English to Farsi with the No Language Left Behind model. Using FNC-1F, we fine-tuned and evaluated three transformer-based models ParsBERT-FA, RoBERTa-fa-ZWNJ, and mDeBERTa-v3. We also validated their performance on real-world, manually curated Farsi test data. Results show that models trained on the translated dataset generalize effectively to native Farsi with only minor performance loss. Beyond presenting a practical approach for building fact-checking datasets in LRLs, this work also establishes the first benchmark for evidence-based fact-checking in Farsi.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.008 | 0.028 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.004 | 0.004 |
| Science and technology studies | 0.002 | 0.002 |
| Scholarly communication | 0.002 | 0.005 |
| Open science | 0.003 | 0.004 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.004 | 0.004 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".