Are Multilingual Models Effective in Code-Switching?

by   Genta Indra Winata, et al.

Multilingual language models have shown decent performance in multilingual and cross-lingual natural language understanding tasks. However, the power of these multilingual models in code-switching tasks has not been fully explored. In this paper, we study the effectiveness of multilingual language models to understand their capability and adaptability to the mixed-language setting by considering the inference speed, performance, and number of parameters to measure their practicality. We conduct experiments in three language pairs on named entity recognition and part-of-speech tagging and compare them with existing methods, such as using bilingual embeddings and multilingual meta-embeddings. Our findings suggest that pre-trained multilingual models do not necessarily guarantee high-quality representations on code-switching, while using meta-embeddings achieves similar results with significantly fewer parameters.


page 1

page 2

page 3

page 4


GLUECoS : An Evaluation Benchmark for Code-Switched NLP

Code-switching is the use of more than one language in the same conversa...

The futility of STILTs for the classification of lexical borrowings in Spanish

The first edition of the IberLEF 2021 shared task on automatic detection...

Hierarchical Meta-Embeddings for Code-Switching Named Entity Recognition

In countries that speak multiple main languages, mixing up different lan...

Call Larisa Ivanovna: Code-Switching Fools Multilingual NLU Models

Practical needs of developing task-oriented dialogue assistants require ...

Knowledge Based Multilingual Language Model

Knowledge enriched language representation learning has shown promising ...

CodeSwitch-Reddit: Exploration of Written Multilingual Discourse in Online Discussion Forums

In contrast to many decades of research on oral code-switching, the stud...

Evaluating Contextualized Language Models for Hungarian

We present an extended comparison of contextualized language models for ...