Previous proposalChallenges & Implications of Deep Learning in Healthcare
Deep Dive Into Building Chat-bots Using Deep Learning
There has been growing interest on shedding boring and cumbersome “search and get thousand results” interface to move towards a “conversational” interface to ease the reception of deluge of information in various web and mobile applications. While a naive search bar that simplified information extraction and delivery of web pages was the rage in early 2000s, exponential increase in data and information on the Internet is making it natural for the search bar to evolve into a smarter and responsive interface. Not only can this search bar be more responsive in terms of guiding users in an iterative aka conversation based search, but it can also be assistive in fetching information that suits the interests of different users. Sooner or later, such conversational interfac will replace the old-generation search bar.
However, building such conversational interface has several technology challenges: a) the interface has to understand and extract intent from natural human inputs in terms of text or speech (this is partly done by the current generation search interfaces), b) it has to generate meaningful dialogue responses to engage users into a conversation, c) it has to understand and use context to assist users in fetching the most relevant information, and while doing so d) make the information results personalized to the user as she uses the interface over time.
How do we solve these technology challenges? Where are we in terms of building systems and solutions to tackle them? Is the ‘conversational search’ a hype or soon to be reality? How are recent advances in deep learning playing their role towards building such ‘chat bots’? In this talk, I will take a deep dive to answer these questions and present the state of the art in building conversational agents using deep learning.
Specifically, I will talk about advances in deep text mining that make it possible to extract intent from text: text segmentation or sequence labeling to extract entities from a sentence or paragraph. I will also talk about, how can one build a dialogue generation system (especially in the absence of a large corpus of dialogues) using information on the web. I will then give an example of a chat bot that uses context in a conversation to recommend apparels and fashion products to consumers, and assists them to arrive at the product that they would like to buy in no time.
Overall, the talk will give an overview of the current state of deep learning techniques to build a chat bot, and provide details a few techniques such as LSTM or sequence to sequence learning in bulding a chat bot in reality.
Motivation: exponential growth in information, limitations of current search interfaces, consumer expectations
Components of building a chat bot: intent extraction, dialogue generation, meaningful search, assistance based conversation
Deep Learning techniques: Text segmentation, Sequence labeling, sentence generation, conversation generation
Demo and example: A chat bot in fashion commerce domain
Vijay Gabale is co-founder and CTO of Infilect, an AI-powered Commerce Platform. Infilect has been building a fashion commerce platform to provide exceptional shopping experiences to the Internet consumers. The company has made several innovations in deep learning to process rich multi-media data (text, image, videos) to improve discovery, search and personalization experiences of online consumers.
Prior to co-founding Infilect, he was a research scientist with IBM Research Labs. He graduated with a PhD from IIT Bombay, India in 2012. He has several top tier research publications and software patents to his name. He is also co-organizer of ‘Deep Learning Bangalore’ meetup. He has been actively working in deep learning for past several years and has give several talks in and outside India on the research and applications of deep learning in e-commerce.