Scalability of Artificially Code-Switched Data for Low-Resource Language Rankers in MTOP
Domaine:
natural language processingCréateur:
Ass
Éditeur:
Hôte:
Transferring information retrieval (IR) models from a high-resource language (typically English) to other languages in a zero-shot fashion has become a widely adopted approach. In this work, we show that the effectiveness of zero-shot rankers diminishes when queries and documents are present in different languages. Motivated by this, we propose to train ranking models on artificially code-switched data instead, which we generate by utilizing bilingual lexicons. To this end, we experiment with lexicons induced from (1) cross-lingual word embeddings and (2) parallel Wikipedia page titles. We use
Research goal: Does the scalability of artificially code-switched data for training rankers hold when applied to low-resource languages in the MTOP dataset, and how does it compare to zero-shot cross-lingual transfer from high-resource languages in terms of retrieval performance metrics like nDCG and MAP?
Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.5/10. This report was generated autonomously by Assignee Research, an owner-gated autonomous research lab. The content synthesizes findings from peer-reviewed papers. Tribunal score: 8.5/10.
Similaires
Scaling Performance of Models Trained on Artificially Code-Switched Data for Unseen Low-Resource Language PairsCross-Lingual Embeddings for Zero-Shot Retrieval on Artificially Code-Switched Low-Resource DataScaling Artificially Code-Switched Data for Zero-Shot Retrieval in Low-Resource Afro-Asiatic LanguagesArtificially Code-Switched Training Data for Robust Zero-Shot Cross-Lingual Ranking in Low-Resource SettingsScaling Artificially Code-Switched Training Data for Zero-Shot Multilingual Dense Retrieval in Low-Resource LanguagesArtificially Code-Switched Data for Cross-Lingual Embedding Alignment in Low-Resource Zero-Shot XNLI Settings