Conversational combines multiple sub-disciplines in AI/ML and NLP, including but not limited to intent classification, entity extraction and linking, state tracking, language generation, question answering (including open domain QA) etc. The field often combines the state-of-the-art in research and industry with new approaches released and quickly being integrated in real world dialogue systems we use in our daily lives. In this talk I will go over some of the key components that go into designing and developing a dialogue system including different taxonomies of conversational systems, key challenges and mitigations, evaluation approaches and then focus specifically on task-oriented dialogue systems which are generally more common in commercial settings.
Hanoz Bhathena is an Applied Machine Learning Scientist Lead at the Machine Learning Center of Excellence at JPMorgan Chase & Co. He has experience executing and leading several data science projects particularly in deep learning, natural language understanding, information extraction and information retrieval. Currently, his focus includes areas like question answering, dialogue systems and semantic search. Previously he worked as a Machine Learning Data Scientist within the Evidence Lab Innovations division at UBS, where he was responsible for developing machine learning models that uncovered insights from unstructured data relevant to investment research. Here, one of his key achievements was the Deep Theme Explorer, an application to conveniently find complex and emerging topics present in large corpora and attribute fine grained sentiment to help predict a company’s exposure to an investment theme. He holds a Master’s degree in Operations Research from Columbia University and a Bachelor’s degree in Electrical Engineering from the University of Mumbai, VJTI. He has also completed the Artificial Intelligence Graduate Certificate program from Stanford University.