Crowdsourcing Diverse Paraphrases for Training Task-oriented Bots

TL;DR

This paper proposes a novel crowdsourcing method to generate syntactically diverse paraphrases, enhancing dataset quality for training task-oriented conversational bots.

Contribution

It introduces a guiding approach to crowd-based paraphrasing that emphasizes syntactic diversity, addressing a gap in existing lexical-focused methods.

Findings

Effective guidance increases syntactic diversity of paraphrases

Improved dataset quality for task-oriented bots

Potential for broader application in NLP data collection

Abstract

A prominent approach to build datasets for training task-oriented bots is crowd-based paraphrasing. Current approaches, however, assume the crowd would naturally provide diverse paraphrases or focus only on lexical diversity. In this WiP we addressed an overlooked aspect of diversity, introducing an approach for guiding the crowdsourcing process towards paraphrases that are syntactically diverse.