Machine Learning Framework for Clinical Text Mining

Lifeng Han, Hao Li, and Haifa Alradhi (Department of Computer Science)

Illustration of different elements of clinical text mining

About the project

The TransformerCRF project team works to develop and share an open-source platform for clinical text mining used by researchers, developers, and clinicians. Their ongoing work aims to benefit the research community through the sharing of toolkits and resources for use in clinical text mining projects, especially in healthcare and life sciences research. Project resources and outcomes, including source code, are openly available on GitHub.

What the Accelerator Fund provided

The Accelerator Fund 2022 provided financial support to assist with two streams of project activity:

  1. Activity One: Creation of open-source multi-tasking learning model using natural language processing technologies for clinical text mining and supporting decision-making.
  2. Activity Two: Creation of human-annotated clinical data for computational models to acquire more knowledge from human expertise.

Full information about all outputs can be found on GitHub.

Impact on research culture

This project leads by example in advocating for open and transparent practices and sharing of code and data in computer science. These contributions will help to improve accuracy and quality of automated information mining, allowing others to benefit from, and build upon the work. Furthermore, the adoption of open principles in the sharing of methodologies, code and data helps to contribute to a lasting shift in the research culture in Manchester. By supporting open ways of working the Office for Open Research hopes that these practices will become embedded in our research culture.