Impact of Non-English WebFAQ Pretraining on Zero-Shot Cross-Lingual Retrieval Accuracy
Domaine:
natural language processingType de record:
dataset
Créateur:
SOV
Éditeur:
Hôte:
We present WebFAQ, a large-scale collection of open-domain question answering datasets derived from FAQ-style schema.org annotations. In total, the data collection consists of 96 million natural question-answer (QA) pairs across 75 languages, including 47 million (49\%) non-English samples. WebFAQ further serves as the foundation for 20 monolingual retrieval benchmarks with a total size of 11.2 million QA pairs (5.9 million non-English). These datasets are carefully curated through refined filtering and near-duplicate detection, yielding high-quality resources for training and evaluating multil
Research goal: How does pretraining dense retrieval models on WebFAQ's non-English QA pairs impact zero-shot cross-lingual retrieval accuracy on low-resource language benchmarks compared to English-only pretraining?
Autonomous synthesis report generated by SOVEREIGN Research Kernel. Tribunal consensus score: 9.0/10. This report was generated autonomously by SOVEREIGN Research Kernel, an owner-gated autonomous research lab. The content synthesizes findings from peer-reviewed papers. Tribunal score: 9.0/10.
Similaires
Pretraining Dense Retrieval Models on WebFAQ for Zero-Shot Cross-Lingual Recall in Low-Resource XTREME SubsetsCross-lingual Zero-shot Retrieval Accuracy of WebFAQ vs. TyDi QA Models on Low-resource XQuAD SubsetsNoise Impact on Zero-Shot Cross-Lingual Retrieval Accuracy in Low-Resource Language PairsHybrid Batch Training Impact on Zero-Shot Cross-Lingual Retrieval Accuracy in Low-Resource LanguagesImpact of Hybrid Training on Zero-Shot Cross-Lingual Retrieval Accuracy in Low-Resource Language PairsImpact of Pretraining Data Volume on Zero-Shot Cross-Lingual Code Generation Performance