Probing for Multilingual Numerical Understanding in Transformer-Based Language Models
This work addresses the challenge of evaluating compositional reasoning in AI models for multilingual contexts, though it is incremental as it extends existing probing methods to numerical data across languages.
The paper tackled the problem of assessing multilingual numerical understanding in transformer-based language models by proposing probing tasks for compositional reasoning over numbers in English, Japanese, Danish, and French, finding that models like DistilBERT, XLM, and BERT perform well on grammaticality judgments but poorly on value comparisons.
Natural language numbers are an example of compositional structures, where larger numbers are composed of operations on smaller numbers. Given that compositional reasoning is a key to natural language understanding, we propose novel multilingual probing tasks tested on DistilBERT, XLM, and BERT to investigate for evidence of compositional reasoning over numerical data in various natural language number systems. By using both grammaticality judgment and value comparison classification tasks in English, Japanese, Danish, and French, we find evidence that the information encoded in these pretrained models' embeddings is sufficient for grammaticality judgments but generally not for value comparisons. We analyze possible reasons for this and discuss how our tasks could be extended in further studies.