Natural Language Processing: How NSF Supports Visionary New Approaches to NLP

Tremendous advances in the capabilities of large language models (LLMs) in recent years have made them a component of many of today’s most exciting technologies, from AI content authoring to deep research information synthesis tools. The evolution of today’s LLMs began with a shift toward data-driven, empirical methods more than 30 years ago. Their astonishing capacities are the result of multiple paradigm shifts over that period, many of which were made possible by the unique nature of US National Science Foundation (NSF) funding. The visionary Natural Language Processing (NLP) research funded through NSF has led to technology with major technical and social impact, in areas including mental health, global information access, education, and the physical sciences. Through individual grants, interdisciplinary convenings and larger special initiatives like AI Institutes, NSF funding has supported transformative research as well as multiple generations of students who then lead new research at universities or enter industry and drive further innovation. These students, whose education combines decades of distilled NLP expertise with their unique talents and perspectives, are perhaps NSF’s most important legacy. The loss of NSF support in any of these areas would threaten the United States’ current, but fragile, global leadership in NLP research and technology for many years to come.