# Cross-lingual transfer learning for spoken language understanding

## TL;DR

This paper introduces a weight transfer method for multilingual spoken language understanding that reduces data requirements and improves performance across languages, demonstrated on ATIS and real-world datasets.

## Contribution

Proposes a simple weight transfer approach for cross-lingual SLU that significantly decreases data needs and enhances monolingual model performance.

## Key findings

- Monolingual models outperform state-of-the-art.
- Data requirements for new language SLU are greatly reduced.
- Multitask training with weight transfer yields better results, with optimal settings varying by module.

## Abstract

Typically, spoken language understanding (SLU) models are trained on annotated data which are costly to gather. Aiming to reduce data needs for bootstrapping a SLU system for a new language, we present a simple but effective weight transfer approach using data from another language. The approach is evaluated with our promising multi-task SLU framework developed towards different languages. We evaluate our approach on the ATIS and a real-world SLU dataset, showing that i) our monolingual models outperform the state-of-the-art, ii) we can reduce data amounts needed for bootstrapping a SLU system for a new language greatly, and iii) while multitask training improves over separate training, different weight transfer settings may work best for different SLU modules.