Multilingual Neural Semantic Parsing for Low-Resourced Languages

TL;DR

This paper presents a multilingual semantic parsing approach for low-resource languages using machine translation and transfer learning, demonstrating improved performance and establishing new baselines on multiple datasets.

Contribution

It introduces a novel multilingual training method leveraging machine translation and pretrained encoders, along with a new dataset for evaluation.

Findings

Joint multilingual training outperforms baselines on TOP dataset.

The approach surpasses state-of-the-art on NLMaps.

Zero-shot performance on Italian reaches 44.9% accuracy.

Abstract

Multilingual semantic parsing is a cost-effective method that allows a single model to understand different languages. However, researchers face a great imbalance of availability of training data, with English being resource rich, and other languages having much less data. To tackle the data limitation problem, we propose using machine translation to bootstrap multilingual training data from the more abundant English data. To compensate for the data quality of machine translated training data, we utilize transfer learning from pretrained multilingual encoders to further improve the model. To evaluate our multilingual models on human-written sentences as opposed to machine translated ones, we introduce a new multilingual semantic parsing dataset in English, Italian and Japanese based on the Facebook Task Oriented Parsing (TOP) dataset. We show that joint multilingual training with pretrained encoders substantially outperforms our baselines on the TOP dataset and outperforms the state-of-the-art model on the public NLMaps dataset. We also establish a new baseline for zero-shot learning on the TOP dataset. We find that a semantic parser trained only on English data achieves a zero-shot performance of 44.9% exact-match accuracy on Italian sentences.