Examining Cross-lingual Contextual Embeddings with Orthogonal Structural Probes
This addresses the challenge of leveraging multilingual embeddings for languages with limited resources, offering a method to improve cross-lingual NLP tasks.
The study investigated whether multilingual contextual embeddings from mBERT encode syntactic and lexical information in a shared cross-lingual space for nine languages, finding that closely related languages require no transformation while others benefit from language-specific orthogonal transformations, leading to successful zero-shot and few-shot cross-lingual parsing.
State-of-the-art contextual embeddings are obtained from large language models available only for a few languages. For others, we need to learn representations using a multilingual model. There is an ongoing debate on whether multilingual embeddings can be aligned in a space shared across many languages. The novel Orthogonal Structural Probe (Limisiewicz and Mareček, 2021) allows us to answer this question for specific linguistic features and learn a projection based only on mono-lingual annotated datasets. We evaluate syntactic (UD) and lexical (WordNet) structural information encoded inmBERT's contextual representations for nine diverse languages. We observe that for languages closely related to English, no transformation is needed. The evaluated information is encoded in a shared cross-lingual embedding space. For other languages, it is beneficial to apply orthogonal transformation learned separately for each language. We successfully apply our findings to zero-shot and few-shot cross-lingual parsing.