Nyberg and CMU team earn top honors in biomedical question-answering challenge

Pittsburgh, PA  |  November 7 — Eric Nyberg, Ph.D and his team from Carnegie Mellon University’s Language Technologies Institute (LTI) at the School of Computer Science earned top honors at BioASQ 2016, a global biomedical semantic indexing and question answering competition. BioASQ challenges include tasks such as hierarchical text classification, machine learning, information retrieval, QA from texts and structured data, multi-document summarization, and others. Top medical and research universities from around the world participated in the annual challenge, including Mayo Clinic, Aristotle University of Thessaloniki and the Hasso-Plattner-Institute at the University of Potsdam, Germany.

The winning team from CMU consisted of Cognistx’s Chief Data Scientist, Eric Nyberg, and two of his students from CMU’s School of Computer Science: Zi Yang, a PhD student at LTI; and Yue Zhou, a student in the Masters of Computational Data Science program. Nyberg, Yang and Zhou received awards for their solutions for Task 4B: Biomedical Semantic QA, which used benchmark datasets containing training and biomedical questions constructed by biomedical experts. The team received first place in the Exact Answers category for batches 3, 4 and 5. For this challenge, Nyberg and his team were given a biomedical question in natural language. The team created a system that automatically produced an accurate, paragraph-sized summary response that was fluent and coherent in natural English language. The team also received second place for their solution to batch 4 in the Concepts category.

The BioASQ challenge started in February 2016, and teams presented their results in August at the BioASQ Workshop at the Association for Computational Linguistics (ACL) Conference at Humboldt University in Berlin. Final results of the challenge were announced October 24, 2016.

See the full article
Learn more about how tasks in BioASQ relate to larger biomedical research efforts