Modeling Personality in a Chatbot
Luka is best known for its conversational agents (chatbots) that mimic a certain personality, e.g. fictional characters from HBO’s Silicon Valley show, Prince and few other celebrities. Nowadays, Luka users can create personal chatbots, their digital avatars based on personal data from social media. Eugenia and Artem will talk about the architecture and a deep learning model underlying these chatbots. This talk discusses recent developments in modeling human text conversation and suggests an approach that combines ranking and a full generative conversational model based on recurrent neural networks.
Artem Rodichev is a Lead Machine Learning Engineer at Luka. He earned his M.Sc. degree in Applied Mathematics and Computer Science from Moscow State University and graduated from Yandex School of Data Analysis. Prior to Luka, Artem accomplished a successful career at Yandex as a software engineer working on NLP-driven search products. Artem’s area of research interest includes Conversational A.I. and Deep Learning techniques for modeling human conversation.