On Enhancing Deep Embedded Clustering for Intent Mining in Goal-Oriented Dialogue Understanding

Discovering user intents plays an indispensable role in natural lan- guage understanding and automated dialogue response. However, labeling intents for new domains from scratch is a daunting process that often requires extensive manual effort from domain experts. To this end, this paper proposes an unsu- pervised approach for discovering intents and automatically producing intention labels from a collection of unlabeled utterances in the context of the banking domain. A proposed two-stage training procedure includes deploying Deep Em- bedded Clustering (DEC), wherein we made significant modifications by using the Sophia optimizer and the Jensen-Shannon divergence measure to simultane- ously learn feature representations and cluster assignments. A set of intent labels for each cluster is then generated by using a dependency parser in the second stage. We empirically show that the proposed unsupervised approach is capable of generating meaningful intent labels and short text clustering while achieving high evaluation scores.

Access paper here