Making Task-Oriented Dialogue Datasets More Natural by Synthetically Generating Indirect User Requests

TL;DR

This paper introduces a method for synthetically generating indirect user requests to enhance task-oriented dialogue datasets, aiming to improve the evaluation and training of smaller models in understanding complex discourse phenomena.

Contribution

It proposes linguistic criteria and an LLM-based pipeline for creating realistic indirect requests, and releases a new dataset for benchmarking model performance on these requests.

Findings

Generated IURs improve model robustness in understanding indirect requests.

The dataset enables better evaluation of smaller models' NLU and DST capabilities.

The pipeline facilitates creating more natural dialogue data with complex discourse features.

Abstract

Indirect User Requests (IURs), such as "It's cold in here" instead of "Could you please increase the temperature?" are common in human-human task-oriented dialogue and require world knowledge and pragmatic reasoning from the listener. While large language models (LLMs) can handle these requests effectively, smaller models deployed on virtual assistants often struggle due to resource constraints. Moreover, existing task-oriented dialogue benchmarks lack sufficient examples of complex discourse phenomena such as indirectness. To address this, we propose a set of linguistic criteria along with an LLM-based pipeline for generating realistic IURs to test natural language understanding (NLU) and dialogue state tracking (DST) models before deployment in a new domain. We also release IndirectRequests, a dataset of IURs based on the Schema Guided Dialog (SGD) corpus, as a comparative testbed for evaluating the performance of smaller models in handling indirect requests.