Analyzing Zero-shot Cross-lingual Transfer in Supervised NLP Tasks
This work addresses the problem of applying NLP models across languages without retraining, providing empirical insights for researchers and practitioners in multilingual NLP, though it is incremental in validating existing hypotheses.
The paper investigated zero-shot cross-lingual transfer in supervised NLP tasks using XLM-RoBERTa, finding that transfer effectiveness varies by task complexity, with semantic textual similarity showing the strongest transfer, sentiment analysis moderate, and machine reading comprehension the weakest.
In zero-shot cross-lingual transfer, a supervised NLP task trained on a corpus in one language is directly applicable to another language without any additional training. A source of cross-lingual transfer can be as straightforward as lexical overlap between languages (e.g., use of the same scripts, shared subwords) that naturally forces text embeddings to occupy a similar representation space. Recently introduced cross-lingual language model (XLM) pretraining brings out neural parameter sharing in Transformer-style networks as the most important factor for the transfer. In this paper, we aim to validate the hypothetically strong cross-lingual transfer properties induced by XLM pretraining. Particularly, we take XLM-RoBERTa (XLMR) in our experiments that extend semantic textual similarity (STS), SQuAD and KorQuAD for machine reading comprehension, sentiment analysis, and alignment of sentence embeddings under various cross-lingual settings. Our results indicate that the presence of cross-lingual transfer is most pronounced in STS, sentiment analysis the next, and MRC the last. That is, the complexity of a downstream task softens the degree of crosslingual transfer. All of our results are empirically observed and measured, and we make our code and data publicly available.