Hannah Zeru

Vol. 1,  No. 1, pp. 1-12, Aug.  2024
10.23246/AAIRJ.2024.01.01.01

Artificial intelligence Natural Language Generation Conversation Grounded Conversation

PDF

This paper explores the landscape of grounded conversation. Current advancements enable language models to engage in natural and informative dialogues with humans, facilitated by diverse conversational datasets designed to enhance the learning of human language. Persona-grounded conversations ensure consistency in model personalities, while knowledge-grounded conversations contribute to the development of intelligent agents capable of retrieving and grounding relevant information. Also, this paper delves into the integration of persona and knowledge datasets, presenting research that aims to make models that are both informative and friendly. Beyond textual interactions, the exploration extends to multi-modality-grounded conversations, ushering in a new era of more comprehensive conversational agents. The paper provides a comprehensive overview of various grounded conversation datasets and training methodologies, encompassing persona, knowledge, persona-knowledge, and multi-modality, to offer insights into the multifaceted dimensions of contemporary language model training.