Ph.D. Thesis Proposal
Nathan J. Blaylock
In order to understand natural language, it is necessary to understand the intentions behind it, (i.e. why an utterance was spoken). Intentions can best be understood by modeling communication as collaboration between agents. Intentions can then be seen as an agent trying to affect the collaboration. Previous work on intention recognition approaches to dialogue has focused only on a small subset of agent collaboration paradigms (master-slave), and thus is unable to account for many discourse phenomena that occur in other paradigms.
We present a preliminary collaborative problem solving model to be used for intention recognition by a conversational agent. This model is not only able to account for previously handled discourse phenomena, but also covers phenomena that occur in various other collaboration paradigms, such as evaluations and interleaving of planning and execution.
We propose, for future research, to complete this model and to build a domain independent intention recognition system based on it for use within the TRIPS dialogue system.