Zero-shot Reading Comprehension by Cross-lingual Transfer Learning with Multi-lingual Language Representation Model
This addresses the problem of limited training data for reading comprehension in many languages, offering a practical solution for multilingual NLP applications.
The paper tackled zero-shot cross-lingual reading comprehension by using a pre-trained multilingual language model, showing that zero-shot transfer is feasible without translating source data, which even degrades performance.
Because it is not feasible to collect training data for every language, there is a growing interest in cross-lingual transfer learning. In this paper, we systematically explore zero-shot cross-lingual transfer learning on reading comprehension tasks with a language representation model pre-trained on multi-lingual corpus. The experimental results show that with pre-trained language representation zero-shot learning is feasible, and translating the source data into the target language is not necessary and even degrades the performance. We further explore what does the model learn in zero-shot setting.