An Ontology for Conversations with Virtual Research Assistants

Conversational artificial intelligence has expanded rapidly in recent years, especially with the growth of large language models (LLMs). Its incorporation in scientific research in the form of research assistants has also become more common-place but remains limited in some capacities, such as in the realm of polymer science. The limitations of LLMs, especially in terms of domain knowledge, warrant the need for other tools, such as knowledge graphs (KGs), to better guide conversations. While such conversational models have been developed in the past, they are generally restricted to particular domains and lack the ability to integrate semantics from various kinds of conversations. Thus, we make progress toward the construction of a universal conversational model that has a focus on the materials domain by combining aspects of existing models. We aim to implement it in such a way that renders it amenable to modifications and usable in a variety of situations. We posit that this model will be adopted and extended by others seeking to accomplish a similar goal in the future.