ACM India Featured Eminent Speaker: Sachindra Joshi

July 29, 2021

The monthly ACM India Featured Eminent Speaker initiative is aimed at highlighting some of our Eminent Speakers who have been doing cutting-edge research in computer science or share an interesting perspective on technology or computing topics. We bring forward these speakers to allow a peek into their work and lives.

The ACM India Eminent Speaker Program (ESP) provides local ACM Professional and Student Chapters in India with direct access to top technology leaders and innovators who will give talks on issues that are important to the computing community. ESP was launched as an ACM India version of ACM's popular Distinguished Speaker Program, in which some India-based speakers also participate.

For the full value of experiencing an interaction with them, sign up to have them give a talk at your event at the website above.

Meet Sachindra Joshi: A Distinguished Engineer at IBM Research-India and our ACM featured speaker of the month.

ACM India: Why is conversational AI so important for the world?

SJ: Conversational AI refers to a set of technologies that facilitate human like interactions between human and computing systems. Over the last few decades computing systems have touched all spheres of our lives, and the way we interact with these systems is going to change tremendously with the application of conversational AI. This would enable greater collaboration across machines and humans and pave the way for significant new automations and increased customer engagement.

ACM India: What is an exciting research problem you solved?

SJ: Significant advances have been made in deep learning based conversational systems. However, their application in real business has been slow and hindered by enormous compute requirements and lack of any control in the way these trained systems interact with end users. We have explored ways in which learnings from the deep learning-based systems can be incorporated in chatbots without huge compute requirements. We have also explored ways in which conversations could be generated using documents and prior conversations, thereby controlling their generation.

ACM India: What are some mentoring tips you would like to share with enthusiasts in this field?

SJ: There is a large gap between the deep learning-based conversation models and the kind of conversation frameworks that industry is using today. There are enormous opportunities in bridging this gap. It requires learning about the advances in deep learning techniques such as transformer networks, large scale pre-trained networks and recent advances on language modeling, as well as learning about more traditional rule-based frameworks utilized by industry today. The best way to build expertise in this area is by getting the experience of building deep learning-based conversation models as well as real chatbots using industry frameworks.

Here is an abstract of one of Sachindra's talks:

Bridging the Gap: Industrial Dialog Framework vs. Deep Learning-based Dialog Systems
There have been tremendous advances in building dialog systems that use past human to human conversation logs and apply deep learning-based techniques on them for learning dialog models. At the same time, several frameworks for building chat-bots have been released by different enterprises such as Google Dialog Flow, IBM Watson Assistant and Microsoft Bot framework. In this talk, I would illustrate the gap that exists in these two parallel threads of work and show how we could apply deep learning-based techniques in the context of industry frameworks to get the best of both worlds.

To learn more about Sachindra, please visit his ESP page.