Unsupervised Flow Discovery from Task-oriented Dialogues

TL;DR

This paper introduces an unsupervised method to automatically discover dialogue flows from task-oriented conversations, reducing manual effort and enabling domain-independent analysis of dialogue structures.

Contribution

It presents a novel approach that clusters dialogue utterances and constructs transition graphs to identify dialogue flows without supervision.

Findings

Successfully extracted meaningful dialogue flows from MultiWOZ dataset.

Introduced an automatic validation metric for flow assessment.

Demonstrated potential for domain-independent flow discovery.

Abstract

The design of dialogue flows is a critical but time-consuming task when developing task-oriented dialogue (TOD) systems. We propose an approach for the unsupervised discovery of flows from dialogue history, thus making the process applicable to any domain for which such an history is available. Briefly, utterances are represented in a vector space and clustered according to their semantic similarity. Clusters, which can be seen as dialogue states, are then used as the vertices of a transition graph for representing the flows visually. We present concrete examples of flows, discovered from MultiWOZ, a public TOD dataset. We further elaborate on their significance and relevance for the underlying conversations and introduce an automatic validation metric for their assessment. Experimental results demonstrate the potential of the proposed approach for extracting meaningful flows from task-oriented conversations.