Few-Shot NLU with Vector Projection Distance and Abstract Triangular CRF

TL;DR

This paper introduces a novel few-shot NLU approach using vector projection distance and an abstract triangular CRF, achieving state-of-the-art results without fine-tuning on target domains.

Contribution

It proposes a new method combining vector projection distance and an abstract triangular CRF to improve few-shot NLU performance across domains.

Findings

Significantly outperforms strong baselines.

Achieves state-of-the-art on Few-Joint and SNIPS benchmarks.

Effective without fine-tuning on target domains.

Abstract

Data sparsity problem is a key challenge of Natural Language Understanding (NLU), especially for a new target domain. By training an NLU model in source domains and applying the model to an arbitrary target domain directly (even without fine-tuning), few-shot NLU becomes crucial to mitigate the data scarcity issue. In this paper, we propose to improve prototypical networks with vector projection distance and abstract triangular Conditional Random Field (CRF) for the few-shot NLU. The vector projection distance exploits projections of contextual word embeddings on label vectors as word-label similarities, which is equivalent to a normalized linear model. The abstract triangular CRF learns domain-agnostic label transitions for joint intent classification and slot filling tasks. Extensive experiments demonstrate that our proposed methods can significantly surpass strong baselines. Specifically, our approach can achieve a new state-of-the-art on two few-shot NLU benchmarks (Few-Joint and SNIPS) in Chinese and English without fine-tuning on target domains.