Natural Language Processing State of Science Conference

HSR&D convened the NLP State of Science Conference on September 9th and 10th, 2015 at the VHA National Conference Center. The purposes of the Conference were to explore new research areas and approaches to accelerate VA research in NLP and to identify balance between innovative NLP research and applied solutions to address a specific problem. This conference brought over 30 leading NLP researchers from non-health care areas, health care plans and the industry, along with leading biomedical NLP researchers from in VA and outside VA. The industry participants included IBM Watson, 3M Health Information System, Kaiser Permanente Health Service Research, Mayo Clinic, and Microsoft Research.

The Conference provided great opportunities for the VA NLP investigators to present their achievements to non-VA NLP experts and federal partners. It also facilitated collaborations between VA and non-VA NLP researchers. The participants expressed excitement about wealth of VA clinical notes that can be used for the NLP research as access to clinical notes for NLP research has been generally challenging. They are interested in conducting innovative NLP research with VA clinical data. Additionally, there were great interest on use of NLP in creating and refining disease phenotypes for genomic research with on-going Million Veterans Program (MVP).

This conference also created opportunities to further engage VA operational partners (e.g. Office of Patient Care Services and Office of Informatics and Analytics) in VA NLP research agenda settings. Partnerships with these Program Offices would be critical in prioritizing uses cases and solving operational grand challenges that the NLP research can provide high values. Successful integration of NLP tools in clinical workflow could enhance human productivity/effectiveness in clinical care delivery and making VHA's health care system closer to learning health care system.

The participants discussed continued methodological challenges associated with biomedical NLP research and opportunities for future research (e.g. NLP research in speech recognition and synthesis, visualization of data, deep learning, and NLP research with social media discourse). Collaborations with academic institutions could provide wonderful training opportunities to post-docs specialized in NLP.

Meeting Materials

• Agenda
• Participant List
  
  
  
• Meeting Intro (David Atkins)
• Key note presentation: "State of Art of Health NLP" (Guergana Savova, Harvard Medical School)
• "NLP research trend in the last five years and VA NLP research accomplishments" (Wendy Chapman, Salt Lake City VAMC)
• Philip Resnik, Noah Smith, and Microsoft presentations
• Panel Discussion: NLP innovations – Healthcare industry perspective on knowledge gaps in NLP research
• Panel Discussion: State of VA NLP research – accomplishments, opportunities, and future challenges
• Panel Discussion: Recommendations for VA NLP research agenda