A Peer-to-Peer Corpus for Conversational Agents for Long-Distance Relationships

Abstract

Recent advances in machine learning, including the development of more effective natural language processing (NLP) models, have enabled the use of text classification and generation algorithms, sentiment and emotion detection models, and intelligent conversational agents, in different domains, from business to healthcare. Specifically, intelligent and conversational agents (e.g., chatbots) are currently incorporated in many applications (e.g., customer care and decision support systems) to automate tasks while simultaneously providing users with a more credible and natural human-like interaction. The availability of NLP corpora is crucial for training conversational agents and increasing their quality and performance. Nevertheless, the availability of domain-specific NLP corpora is crucial for training conversational agents, especially in applications that focus on mental health counseling and support. In this paper, we introduce a corpus especially designed for NLP tasks that focus on providing bi-national couples in a long-term relationship with mental health support. Our dataset contains over 4000 posts and users’ reactions published on social media groups dealing with COVID-19 travel restrictions. We detail the content of the dataset, its format, and its use in the development of NLP applications.