Conversational agents, commonly known as chatbots are computer programs that engage in conversation using Natural Language (NL) dialogue with a human user. NL is highly ambiguous; it is discrete, symbolic and highly variable. For example, the utterance (1) “I ate a pizza with friends” and the utterance (2) ‘I ate a pizza with olives’ demonstrates this high variability, and the core message in utterance (1) can as be expressed as ‘friends and I shared a pizza’. Natural language is functional, and from this utterance we form mental representations.
Conversational agent categories are interactive (a) FAQs; (b) form filling (c), question answering (QA); (d)NL interface for databases [1]. Finally, our focus is dialogue planning in conversation which may use a-d. Dialogue systems involve dialogue management as part of the conversation to achieve a goal (as part of the social interaction) in a domain specific knowledge environment. Dialogue systems are based on multi-turn conversations, story comprehension, natural language understanding and human cognition [2-4].
Since 2018 the NLP landscape has experienced the move from machine learning NLP problems of shallow models, to neural networks and NLP language model, demonstrating superior results on many NLP tasks, such as advanced QA, language inference, translation, and sentence prediction [5-7]. However, as NL is inherent, ambiguous, complex, and dynamic in nature – we continue to experience this long-standing issue of refining the accuracy of the interpretation of meaning to provide a realistic dialogue to support the human-to-computer communication [8-11].
This project aims to design and develop a hybrid [2, 3, 12-14] – linguistic and machine learning solution for a conversational – dialogue system, in a specific domain. This research project will sit in our AI and Visual Computing Research Unit, part of the AI Research group [15]. Here we have a small research group in NLP who has published work on a linguistically motivated text based conversational agent and part of a globally established NLP, and knowledge representation community. This project will be underpinned by a linguistic engine backbone and form part of new framework that utilise aspects of the latest NLP neural, language models. This research project will serve as blueprint of a novel conversational AI. The research project will have much impact in the need for bespoke conversational support systems for different domains. One potential use case to be explored is in healthcare.
Students who have a background in natural language processing and an interest in language itself - with expertise in any of the following languages (Python, Java, C, C++, or C#), and are interested in human computer interaction and integration development - are welcome. The domain of this project can be adjusted as per the qualification and interests of students. For further details, please enquire.