AI language processing for disambiguation and clarification of intent
Project Leader: Adrian Pearce
Primary Contact: Adrian Pearce (firstname.lastname@example.org)
Keywords: artificial intelligence; autonomous systems; computational linguistics; machine learning
Disciplines: Computing and Information Systems
This project is motivated by the overarching need to clarify the intent of queries encountered in dialogue systems. At the core of this is the interplay between disambiguation of meaning, given the current context, and establishing intent. A solution to this challenge will need to tackle the integration of statistical language processing with logical reasoning behind intent. In this thesis, natural language processing will be focused on a domain-specific interpretation. Relevant applications include better ways for briefing and debriefing pilots during training exercises. This will include developing a language for pilots, ie, a dialogue driven scenario involving pilots talking to each other, and the AI, to establish what they are doing. At the core of classification of intent, is: what is the language of pilots? The learning modalities relevant to this thesis includes human in the loop reinforcement learning to learn the initial ontology, to correct and up-date the ontology, and to manage word embedding and getting context right.
The project is presently recruiting PhD candidates. The successful candidate will be working with a team of academics and postdocs. You will join a vibrant AI research group, including award winning researchers in fields of Automated Planning and Scheduling, and a group producing PhD's winning Best Dissertation Awards (2018, 2016, 2015). The group has significant supporting AI infrastructure, including the Lapkt AI planning toolkit (lapkt.org), and the PDDL planning domains (solver.planning.domains) developed in conjunction with The University of Melbourne, MIT, Kings College London and ANU.