Noise in Voice a" Evaluation of Speech Recognition and Information Extraction in Hospitals
Hanna Suominen
NICTA SML SEMINARDATE: 2013-05-30
TIME: 11:15:00 - 12:15:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
During this talk, we focus on technological challenges of noise in voice in the context of speech recognition and information extraction. First, we use as motivation the task of using speech recognition to transcribe verbal nursing handover information and information extraction to populate a structured form with this text. Second, we divide the challenges into the categories of noise in language, background, and erroneous transcriptions and review existing technologies and their performance in these noisy settings. Third, we describe our experimental results with data from Australian handovers. This includes new evaluation datasets; comparisons of microphones and recorders; accuracy assessments in transcription and form-filling; and a software demonstration. Finally, we welcome discussion on further analyses.
BIO:
Dr Hanna Suominen works as a machine learning researcher in the Human Performance Improvement project at NICTA Canberra Research Laboratory, Australia. She also holds an adjunct position in the Australian National University, College of Engineering and Computer Science, Canberra, Australia and is a post doc researcher of the Academy of Finland. Dr Hanna Suominen has a strong expertise in health care services and information systems. Her PhD dissertation on machine learning and clinical text has honours of belonging to the 10 % elite of the field internationally. She has coordinated a Finnish consortium for developing health information and language technology as well as contributed to similar international research networks. She has worked in the Finnish national positron emission centre and taught numerous university courses on bioinformatics and information systems in healthcare. Her other research merits include approximately 35 scientific publications, two best paper awarded contributions, the third prize in the International Medical Natural Language Processing Challenge 2007, Venture Cup 2007-2008 award further funding for research commercialization, and several personal and collaborative research grants (e.g., Academy of Finland, Nordforsk, Tekes). Her language skills include excellent or native English, Finnish and Swedish. Please do not hesitate to contact Dr Suominen for research collaboration or MSc/PhD projects. See http://www.nicta.com.au/people/hanna_suominen for further details and student projects.
