Revisiting Mahalanobis Distance for Transformer-Based Out-of-Domain Detection

TL;DR

This paper demonstrates that fine-tuning Transformer encoders combined with Mahalanobis distance significantly improves out-of-domain detection in intent classification, setting new state-of-the-art results without requiring out-of-domain training data.

Contribution

It introduces a novel approach using fine-tuned Transformer representations and Mahalanobis distance for out-of-domain detection, outperforming existing methods.

Findings

Mahalanobis distance with Transformer embeddings outperforms other methods.

Fine-tuning Transformer encoders improves in-domain representation homogeneity.

Proposed method achieves new state-of-the-art results on three datasets.

Abstract

Real-life applications, heavily relying on machine learning, such as dialog systems, demand out-of-domain detection methods. Intent classification models should be equipped with a mechanism to distinguish seen intents from unseen ones so that the dialog agent is capable of rejecting the latter and avoiding undesired behavior. However, despite increasing attention paid to the task, the best practices for out-of-domain intent detection have not yet been fully established. This paper conducts a thorough comparison of out-of-domain intent detection methods. We prioritize the methods, not requiring access to out-of-domain data during training, gathering of which is extremely time- and labor-consuming due to lexical and stylistic variation of user utterances. We evaluate multiple contextual encoders and methods, proven to be efficient, on three standard datasets for intent classification, expanded with out-of-domain utterances. Our main findings show that fine-tuning Transformer-based encoders on in-domain data leads to superior results. Mahalanobis distance, together with utterance representations, derived from Transformer-based encoders, outperforms other methods by a wide margin and establishes new state-of-the-art results for all datasets. The broader analysis shows that the reason for success lies in the fact that the fine-tuned Transformer is capable of constructing homogeneous representations of in-domain utterances, revealing geometrical disparity to out of domain utterances. In turn, the Mahalanobis distance captures this disparity easily. The code is available in our GitHub repo: https://github.com/huawei-noah/noah-research/tree/master/Maha_OOD .