Joint Learning of Word and Label Embeddings for Sequence Labelling in Spoken Language Understanding

TL;DR

This paper introduces a novel architecture for slot filling in spoken language understanding that jointly learns word and label embeddings, achieving state-of-the-art results with fewer parameters by avoiding contextual windows.

Contribution

It presents a label embedding method that does not require label embeddings as input and computes contextual distances to improve efficiency and performance.

Findings

Achieves state-of-the-art performance on spoken dialogue datasets.

Uses fewer trainable parameters than existing methods.

Reduces memory footprint by avoiding contextual windows.

Abstract

We propose an architecture to jointly learn word and label embeddings for slot filling in spoken language understanding. The proposed approach encodes labels using a combination of word embeddings and straightforward word-label association from the training data. Compared to the state-of-the-art methods, our approach does not require label embeddings as part of the input and therefore lends itself nicely to a wide range of model architectures. In addition, our architecture computes contextual distances between words and labels to avoid adding contextual windows, thus reducing memory footprint. We validate the approach on established spoken dialogue datasets and show that it can achieve state-of-the-art performance with much fewer trainable parameters.