It is now my fourth and final year studying MEng Artificial Intelligence at the University of Reading. For my final year project I shall be developing a system that follows the thread (or context) of a conversation between discrete items of textual input. This is called Context Aware A.I.
Background research into the subject has revealed that most context-aware systems define context as data related to the user's physical locale such as location, local wifi networks, phone signal strength etc. This means that my research is coming at this problem from a different angle as I am using context to refer to the context of a conversation.
From my initial reading I plan to attempt to tackle this issue by creating ontological categories for every word inputted to the system. This should allow it to quickly determin which words are important for understanding and context and ignore words which are less important to understanding the context. As there is not time to teach the system the entire dictionary, a dictionary API will be used as a data-source to learn new words as they are entered into the system, and will be used to asign their ontology category to them.
After the keywords for contextual information have been detected, these will be used to determin if the context of this input is the same as the previous, and, if so the meaning will be incorportated into the previous understanding. This means the system will be able to follow context through discrete inputs. Furthermore, it is planned that past contextual information is not removed from the system, such that the system can later pick-up the thread of a previous topic and thus give a far more natural response to conversation.
As I work on this project and develop it, I will continue to write new posts about it.