Biomedical Text Processing

For biomedical researchers and clinicians, keeping up-to-date with the latest scientific research findings is becoming a laborious and an increasingly intractable task. Human scientific knowledge is growing at an explosive pace and the world's output of scientific literature is increasing dramatically. One new article is added to the medical literature every 26 seconds or less.

CHI is working to develop new computational methods and tools to mine, extract, and synthesise medical and biological knowledge across heterogeneous sources using natural language processing techniques. Our projects involve investigating algorithms that use machine learning techniques to automatically extract knowledge customised to the user. Structured knowledge gathered from text mining techniques can enable automated meta-analyses and summarisation of relevant information.

Information Extraction and Text Summarisation

Randomised Controlled Trials are a primary source of evidence used by practitioners and clinicians to guide and inform decisions as well as research directions. The application of computational retrieval and summarisation methods to this continually expanding class of literature carries many benefits. Knowledge of current best practices and new discoveries is kept up-to-date, relevant evidence becomes more readily accesible, and decision-making improves. These benefits ultimately contribute to improved patient outcomes through better health practices.

Biological Event Extraction

The automatic identification and extraction of bio-molecular events is a growing area of interest within the domain of biomedical natural language processing. Extraction of related events plays an important role in the discovery of causal pathways that hold the key to unlocking the processes by which diseases and infections occur. Bio-event extraction is an information retrieval task that heavily relies on the recognition of named entities such as proteins and genes. Molecular events related to these entities are then identified through the use of natural language processing methods such as syntactic parsing and semantic processing.