Conversational dataset reveals how tone changes during chats
ConvoDrift: A Multi-Turn Conversational Dataset for Modeling Stylistic Tone Evolution
Computation and LanguageArtificial Intelligence
Summary
People's way of speaking changes during conversations, but most language datasets treat style as the same throughout. The authors created ConvoDrift, a dataset that tracks how tone shifts across multiple turns while the meaning stays the same. They tested it carefully with humans and computers, showing that style changes affect word choices without altering the core message. This helps computers understand and model how people change their talking style over time.
What this means in practice
- •For chatbot developers: Use the dataset to improve chatbots by modeling how conversational tone naturally shifts over multiple turns while keeping meaning consistent.
- •For customer service teams: Develop systems that personalize responses by gradually adapting style during multi-turn conversations with customers.
Authors
Vihindi Kotalawala, Pamoda Dilranga, Gayani Thoradeniya, Prasan Yapa
Abstract
The evolution of linguistic style in conversations is an underexplored issue in NLP. Most style-control datasets focus on sentences or assume a static style throughout, missing the dynamic shifts that occur as user preferences change during interactions. We introduce ConvoDrift, a dataset designed to model progressive stylistic conversational tone drift under fixed semantic intent. It is built on 15,727 shared multi-turn conversational structures for adaptation and persona-conditioned alignment methods. It consists of six prompt-response pairs per conversation, each with the annotation of style drift and style direction labels. These pairs cover a range of communication genres. We further derive a complementary pairwise dataset by pairing semantically equivalent but stylistically distinct responses and annotating persona-conditioned preferences using five distinct style communication personas, enabling the controlled study of personalisation and pluralistic alignment in language tone. In addition to dataset construction, we conduct a comprehensive evaluation involving human validation, LLM-as-judge assessment, and automatic lexical and semantic evaluations. Across seven Likert criteria annotated by three human annotators, the average Krippendorff's alpha is 0.88, and our lexical and semantic analyses show that drift events induce lexical changes while preserving semantic similarity.