University of Michigan hosts NLP @ Michigan Day

On Tuesday, March 11, the University of Michigan hosted NLP @ Michigan Day, a daylong symposium focused on the latest developments in natural language processing (NLP). This rapidly evolving subfield of artificial intelligence (AI) centers on enabling computers to understand, interpret, and generate human language, including through advances in large language models (LLMs). Over 100 researchers and practitioners gathered at the Rackham Graduate School to explore cutting-edge advances and foster collaborations within the NLP community.

Organized by the Michigan AI Lab and chaired by PhD students Muhammad Khalifa and Siyang Liu, the event featured two engaging keynote talks, an interactive poster session, and roundtable discussions. It provided a venue for NLP researchers to exchange ideas, learn about the latest findings, and build connections with fellow experts.

Prof. Parisa Kordjamshidi delivered a keynote titled “Compositional Reasoning for Natural Language Comprehension and Grounding Leveraging Neuro-Symbolic AI.”

Prof. Mohit Bansal delivered a keynote titled “Trustworthy Planning Agents for Collaborative Reasoning and Multimodal Generation.”

Attendees presented their research in NLP at the afternoon poster session.

Attendees presented their research in NLP at the afternoon poster session.

Prof. Rada Mihalcea discusses NLP topics with attendees at the roundtable lunch.

The symposium began with a keynote presentation by Parisa Kordjamshidi, associate professor of computer science and engineering at Michigan State University, titled “Compositional Reasoning for Natural Language Comprehension and Grounding Leveraging Neuro-Symbolic AI.” In her talk, Kordjamshidi discussed the challenges and opportunities in using symbolic representations to enhance the reasoning capabilities of large language models (LLMs), and how they can be leveraged to improve language comprehension and grounding in real-world situations.

Attendees then participated in a poster session showcasing diverse projects being done in the NLP space. The presentations offered a glimpse into dynamic landscape of NLP research, allowing attendees to discuss and gain feedback on their work and learn about new topics and applications in this area.

Over lunch, participants engaged in a roundtable discussion mixer, promoting deeper conversations and networking around key topics within NLP and the broader field of AI, from AI safety and ethics to NLP in medicine and healthcare.

The afternoon keynote was delivered by Mohit Bansal, John R. & Louise S. Parker Distinguished Professor at the University of North Carolina at Chapel Hill, and was titled “Trustworthy Planning Agents for Collaborative Reasoning and Multimodal Generation.” Bansal shared his latest work in developing AI planning agents capable of robust multi-agent reasoning and multimodal generation, emphasizing the importance of trustworthiness and adaptivity in AI interactions.

The event concluded with closing remarks and a recognition of the outstanding contributions of the day’s participants, including the winners of the symposium’s research awards. The Best Research Impact and Innovation Award was presented to Do June Min, a PhD student in CSE at U-M, for his project titled “Speech Retrieval-Augmented Generation without Automatic Speech Recognition.” The Best Presentation Award was given to Ayoung Lee, also a CSE PhD student, for her project titled “Can LLMs Make Value Judgments from Various Perspectives?d

“With the advent of LLMs, there is ever-increasing excitement about NLP and its potential applications,” said Rada Mihalcea, Janice M. Jenkins Collegiate Professor of Computer Science and Engineering and director of the Michigan AI Lab. “This symposium provided a fantastic opportunity to learn about pioneering research being done in this area and form connections across the NLP research community.”