Words of Wisdom From the 2019 Innovation Challenge Winners – PILOT

We are accepting applications for the 2020 PILOT Innovation Challenge from teams who can help broadcasters improve real-time decentralized collaboration. We asked last year’s winners to reflect on their experiences and provide words of wisdom for this year’s applicants. One team called it “a perfect opportunity for a graduate student.”

Last year’s challenge asked teams of researchers to build an AI character that could have conversations with individual viewers, listeners or consumers. PILOT allocated the $150,000 prize between two teams with different takes on the theme. A team from Michigan State University’s NextGen Media Innovation Lab and the Integrated Pattern Recognition & Biometrics Lab developed DeepTalk: A Conversational Agent for Broadcasters. DeepTalk is a conversational agent, like Siri, that can be trained through deep learning to deliver news in the voice of a local broadcaster. Meanwhile, a team from the University of Minnesota’s Department of Computer Science and Engineering developed Jukebot, a chatbot API capable of answering simple questions and getting feedback from radio station listeners.

Both teams took the time to answer some questions about the Innovation Challenge and offer some advice and wisdom for teams thinking about applying to the 2020 Innovation Challenge. The Michigan State University team said they would “highly recommend researchers to consider this opportunity since their contribution can have a tangible impact in the field of broadcasting.”

Watch the video from the University of Minnesota Team, or read on for a complete Q&A.

University of Minnesota (UMN): The Innovation Challenge has helped us by expanding our knowledge of NLP and chatbots and it has been a great practice for our teamwork and programming skills. Watching our chatbot get more and more features has been very interesting and has made us excited to take on more computer science challenges in the future.

Michigan State University (MSU): We are developing an AI-based conversational agent as part of this challenge. The Innovation Challenge has helped us work on cutting edge technology that can benefit not only the broadcasting industry but also other types of voice-driven applications, including digital personal assistants such as Siri, Cortana, Alexa and Google Assistant. After this project is complete, we plan to study the impact of AI on journalism and broadcasting. Using the conversational agent developed through this project, we plan to advance a research agenda that investigates audience perception of the use of AI agents in news broadcasts and delivery.

UMN: The team met twice a week to plan our thoughts and ideas and update each other on progress. As students, we got engaged by getting an insight into advanced projects with deadlines and practical learning of industry standard tools, while as researchers we got the freedom to lead and design tasks while working with the dynamic needs of a company.

MSU: The program brought together two different research groups in Michigan State University – one from the College of Communication Arts and Sciences and another from the College of Engineering. In addition, the involvement of WKAR helped us obtain realistic data for conducting this research. The ensuing collaboration allowed us to approach the problem from both a technological perspective as well as a societal impact perspective. Further, the program engaged one PhD student who looked into both these aspects when developing the solution. Portions of the work undertaken in the Innovation Challenge will be included in the student’s PhD dissertation.

UMN: We would definitely recommend other researchers to apply if given the chance. It has provided us a great opportunity to develop a project we may not have had the resources to otherwise work on, and all that you risk by applying is getting experience in writing a proposal.

MSU: The Innovation Challenge allows researchers to consider innovative applications of their technology. The field of broadcasting has evolved by leaps and bounds over the past decade due to the widespread availability of the internet, social media applications and computational resources. We would highly recommend researchers to consider this opportunity since their contribution can have a tangible impact in the field of broadcasting.

UMN: I think it’s definitely worth one’s time to apply and enter this competition. The end result shouldn’t be the focus; the journey towards it should. People will get valuable experience writing proposals in a competitive environment and charting out the timeline for the project. Even if you don’t think your proposal is good enough, you should just submit it. It’s a good opportunity that would be wasted otherwise.

MSU: Potential applicants must have a good knowledge of the broadcasting industry and understand the needs of the consumer. Further, they must be able to articulate the practical impact of the technology they are developing. They must also take into account the perception of the end customer (i.e., the listener) and the societal implications of their technology (e.g., issues related to privacy and ethics).

UMN: The competition has been a perfect opportunity for a graduate student to get the full experience of the entire process of doing a project, going from the initial step to develop the idea into a proposal, apply for funding, find students to work on the project, manage the group to keep everyone engaged and productive and ensure progress towards the final delivery of the results. This is rarely available in academia, since students generally cannot submit proposals. The project has created friendship among the students who have been building the system together and feel responsible for doing their share of the work. Of course, the funding we received helped keep everyone committed.

MSU: It was exciting to create technology that can benefit the broadcasting industry while keeping the consumer in mind. Creating a conversational agent is not an easy task. It required a careful study of multiple topics including natural language processing, speech synthesis and deep learning-based style transfer. The collaboration between different research groups further enhanced the experience of working on the challenge.