Abstract: Patient's testimonials provide valuable insights to define and characterize Quality of life and patient's perspective of their disease, whether it is about their symptoms, or their advice to others, or what makes them suffer. Extracting information from these stories is therefore relevant to improving care. However, traditional NLP methods of topic modelling are not adapted to these generally short health-specific texts, in majority because they are based on the co-occurrence of words. We propose here a short introduction of the state of the art of these methods and a solution adapted to the extraction of relevant information from this type of textual data.
I: State of the art on Topic Modelling
II: Use cases in health textual data
III: Analytical pipeline demonstration
Use of python and basic NLP
Bio: Mélissa is a data scientist engineer. Over the past 7 years working at Quinten Health in the healthcare sector as a Project Manager in data science, she has participated in the development of several decision support solutions powered by AI, e.g. for rare disease diagnosis, disease progression modelling and endotyping, or evaluation of population heterogeneity. She leds multiple studies of real-world data using advanced analytics methods to characterize phenotypes and disease progression for neurological conditions, cardiovascular diseases, and oncology, for pharma companies, research organizations and care providers. Currently, she is managing at Quinten Health the development of AI-powered solutions to support R&D decisions using RW data for our client.