Patient Opinion Mining from Feedback Comments

This project explores how far machine learning can automatically classify patient comments into key topics - staff attitude, care quality, environment and waiting time, among others. I present an annotated gold-standard corpus developed in two NHS settings in the UK (a general hospital and a mental health trust) and two modelling approaches based on segment- and comment-level representations. The models use various machine learning methods and rely on n-gram features.

The corpus and methods were published as part of the DEPEND mixed-methods study (Health and Social Care Delivery Research, 2020).