The iSchools organization and 2021 Doctoral Dissertation Award Chairs have named alumnus Souvick Ghosh, Ph.D.‘20, Runner Up for the 2021 iSchools Doctoral Dissertation Award for his dissertation “Exploring Intelligent Functionalities of Spoken Conversational Search Systems.” The award chairs wrote, in praise of Ghosh’s dissertation, “the thesis combines the use of developing methods, an emerging technology, and is also true to ‘the best traditions of our field.’”
Ghosh, an assistant professor and the director of the Intelligent Conversational Agents and Neural Networks (ICANN) Lab at the San Jose State University School of Information, said, “I am honored to be a runner-up for this award. There are so many outstanding dissertations from across iSchools members and the award is a recognition of the quality and importance of the research which I conducted as part of my dissertation. It gives me the platform to do extend the work and implement it for social good.”
He will receive the award during iConference 2021, which will take place virtually from March 28-31, 2021. The award includes a $1,000 runner-up prize that Ghosh said he plans on investing in software and computational resources for his lab.
I am honored to be a runner-up for this award. There are so many outstanding dissertations from across iSchools members and the award is a recognition of the quality and importance of the research which I conducted as part of my dissertation. It gives me the platform to do extend the work and implement it for social good.
Ghosh was nominated for the award by the director of SC&I’s Ph.D. Program and Professor of Communication Jennifer Theiss. His dissertation work was completed under the guidance of his dissertation committee comprised of Chirag Shah, a former faculty member at SC&I, who is now an associate professor at the Information School at the University of Washington in Seattle; Distinguished Professor Emeritus of Library and Information Science Nicholas Belkin; Assistant Professor of Communication Katherine Ognyanova, and Vanessa Murdock, Ph.D., at Amazon Research.
According to the iSchools Organization website, “The Schools Doctoral Dissertation Award is an annual competition recognizing the year’s most outstanding dissertations from across iSchools membership. Introduced in 2013, award nominations are solicited from all member schools and subjected to a thorough review by the Doctoral Dissertation Committee. The 2021 Award selection process was chaired by George Buchanan of the University of Melbourne iSchool and Udo Kruschwitz of the iSchool at Universität Regensburg.”
In an announcement posted to the ischools website, award committee members explained the reasons they chose Ghosh for the award. “This thesis was recommended and highly rated throughout the review process, and it addresses an increasingly common new technology. The judges applauded the approach taken saying it ‘uses many diverse methods,’ combines ‘reliable and well-considered use of statistics’ and, vitally, possesses a ‘stress on reliability.’”
Describing how the award will enhance his future research, Ghosh it will help him advance his goals as the ICANN lab director. “The award is a recognition of the research which I accomplished as a Ph.D. student. Many of the ongoing projects at the lab are an extension of my dissertation work. The award gives me the motivation to work harder and highlights the importance of this line of research. I hope that the award helps me attract new graduate students to my lab and motivate the existing ones.”
My research objective is to improve the existing state-of-the-art voice-based conversational systems by making them more human-like, user-friendly, and accessible.
Describing his doctoral and ongoing research, Ghosh said, “Existing state-of-the-art spoken conversational search systems - for example, Amazon Alexa, Apple Siri, or Google Assistant - are good for simple search tasks (‘Who is the President of the United States?’) or smart home tasks (‘Turn off the study lights’). However, these systems often fail to recognize the information needs of the user, especially for complex search tasks where the question is not so straightforward and requires analyzing and evaluating results (for example, searching for the right perfume or camera).
“My dissertation explores strategies to facilitate user-system communication and better identify the user’s search intent. I use human-computer interaction, natural language processing, and deep learning techniques to analyze the human- and system-level understanding of discourse in user-system information-seeking conversations. My research objective is to improve the existing state-of-the-art voice-based conversational systems by making them more human-like, user-friendly, and accessible.
“The findings from my study could be applied to future designs of voice-based assistants in libraries and museums, in medical and healthcare domains, or for people who are visually challenged. If we implement the proposed improvements to a conversational system, the system should allow the user to backtrack, evaluate results, and implement alternate search strategies. The system-level feedback should also help the user pick the source of information, evaluate the trustworthiness of the result, and understand the limitations of the system.
“There are two ways of implementing the recommendations. My goal is to create an open-sourced project to develop a voice-based assistant (a mobile application) which could then be installed on any mobile device. The other option would be teaming up with any of the industry giants and test it on their system.”
Ghosh’s Ph.D. advisor, Chirag Shah, said, “Conversational systems are becoming enormously popular and useful in all kinds of contexts — from e-commerce and entertainment to education and healthcare. These systems are helping very young kids to interact with information before they could read or write; they are helping seniors to use informational services in an easy, natural way. Souvick’s dissertation focuses on such systems with a specific purpose of understanding the user’s speech or dialogue acts associated with their utterances. His work shows not only how to do this better, but also when to do a follow-up to get a clarification from the user. With a unique combination of methods from information science, communication, and human-computer interaction (HCI), Souvick has successfully and effectively demonstrated a logical next step we need to take for developing next generation conversational systems. I am very pleased to see the iConference recognizing the importance of this work by bestowing this honor to Souvick’s dissertation.”
Souvick has successfully and effectively demonstrated a logical next step we need to take for developing next generation conversational systems. I am very pleased to see the iConference recognizing the importance of this work by bestowing this honor to Souvick’s dissertation.
Explaining how pursuing his Ph.D. at SC&I has helped him succeed, Ghosh said, “My Ph.D. has been a wonderful journey for me, and I found my intellectual home at Rutgers SC&I. Coming from a computer science background, SC&I helped me transition from an engineer to a researcher. SC&I allowed me to become more responsible socially and a better teacher overall. My advisor, Dr. Shah, has been the friend, philosopher, and guide that any doctoral student would be lucky to have. I am especially thankful to Professor Belkin, who has always been a constant source of inspiration and knowledge. I had a fantastic dissertation committee, amazingly supportive Program Directors (Dr. Radford and Dr. Theiss) and the SC&I staff, and my friends from InfoSeeking, and SC&I DSA.”
To read more biographical information or to see a copy of Ghosh’s dissertation, click on the iSchools website here.