Generating Hard-Negative Out-of-Scope Data with ChatGPT for Intent Classification

TL;DR

This paper introduces a method using ChatGPT to generate hard-negative out-of-scope data for intent classification, demonstrating that training with this data enhances classifier robustness against challenging out-of-scope utterances.

Contribution

The paper presents an automated, cost-effective technique to generate hard-negative OOS data with ChatGPT and evaluates its impact on improving intent classifier robustness.

Findings

Classifiers struggle more with hard-negative OOS utterances.

Training with hard-negative OOS data improves classifier robustness.

The technique enables easy creation of challenging OOS datasets.

Abstract

Intent classifiers must be able to distinguish when a user's utterance does not belong to any supported intent to avoid producing incorrect and unrelated system responses. Although out-of-scope (OOS) detection for intent classifiers has been studied, previous work has not yet studied changes in classifier performance against hard-negative out-of-scope utterances (i.e., inputs that share common features with in-scope data, but are actually out-of-scope). We present an automated technique to generate hard-negative OOS data using ChatGPT. We use our technique to build five new hard-negative OOS datasets, and evaluate each against three benchmark intent classifiers. We show that classifiers struggle to correctly identify hard-negative OOS utterances more than general OOS utterances. Finally, we show that incorporating hard-negative OOS data for training improves model robustness when detecting hard-negative OOS data and general OOS data. Our technique, datasets, and evaluation address an important void in the field, offering a straightforward and inexpensive way to collect hard-negative OOS data and improve intent classifiers' robustness.