Relying on Discourse Analysis to Manage Dialogues for a Chatbot

Chatbots are becoming fairly popular, however, open-source chatbot systems are lagging behind. We introduce a platform for transactional and question-answering chatbot based on machine learning and linguistic analysis of OpenNLP for search engineers and generalists. We will learn how to design a dialogue manager for a given domain as well as how to populate a chatbot with knowledge. The audience will learn how to apply statistical and inductive machine learning to syntactic and rhetoric parsing data. A special focus will be on how to leverage discourse analysis.

We will assume that the audience has a basic knowledge of deep learning and machine learning concepts, such as loss function, model training, model inference, and feed-forward conv. neural networks.

Speaker: Boris Galitsky, Oracle