Leveraging External Knowledge for Out-Of-Vocabulary Entity Labeling

TL;DR

This paper introduces a neural network that uses external knowledge bases to improve the classification of unseen slot keys and values in multi-domain dialogue state tracking, significantly enhancing model performance.

Contribution

It proposes a novel approach leveraging external KBs to generate candidate slot keys and values for out-of-vocabulary scenarios, outperforming previous methods relying on predefined mappings.

Findings

57.7% increase in F1 score

82.7% increase in accuracy

Effective handling of unseen slots in dialogue systems

Abstract

Dealing with previously unseen slots is a challenging problem in a real-world multi-domain dialogue state tracking task. Other approaches rely on predefined mappings to generate candidate slot keys, as well as their associated values. This, however, may fail when the key, the value, or both, are not seen during training. To address this problem we introduce a neural network that leverages external knowledge bases (KBs) to better classify out-of-vocabulary slot keys and values. This network projects the slot into an attribute space derived from the KB, and, by leveraging similarities in this space, we propose candidate slot keys and values to the dialogue state tracker. We provide extensive experiments that demonstrate that our stratagem can improve upon a previous approach, which relies on predefined candidate mappings. In particular, we evaluate this approach by training a state-of-the-art model with candidates generated from our network, and obtained relative increases of 57.7% and 82.7% in F1 score and accuracy, respectively, for the aforementioned model, when compared to the current candidate generation strategy.