Skip to main content

Semi automated systematic reviews

Research Team: Dr. Leo Ng, Mr. Peter Edwards
CIC Specialist: Dr. Kevin Chai

Researchers spend many hours reviewing literature to synthesise scientific knowledge and evidence. A systematic review uses rigorous, documented and reproducible methods to identify, select and critically appraise all relevant research for a clearly formulated research question. These reviews are performed in analysing health and medical studies (controlled trials) and are considered the highest level of evidence in evaluating the effectiveness of healthcare interventions and informing recommendations. However, a systematic review article can take over 60 weeks to publish with the majority of time spent in screening thousands of articles for relevance.

Dr. Leo Ng from the School of Physiotherapy and Exercise Science in collaboration with Dr. Kevin Chai from the CIC, explored using machine learning and natural language processing techniques to develop an algorithm that could semi-automate the article screening process. Data related to the screening process of completed systematic reviews was collected by Leo from colleagues to validate the developed algorithm. i.e. how quickly can the algorithm help researchers identify the final set of relevant articles used for a given systematic review. Preliminary experiments show that the algorithm was able to reduce the screening time by more than 80% (e.g. only 200 abstracts would need to be reviewed from an initial set of 1,000), which can significantly reduce the overall time required to publish a systematic review article.

The project has received interest and won the inaugural EduGrowth LaunchPad Business Start EdTech competition receiving a cash prize of $5,000 AUD and mentorship for potential commercialisation. Peter Edwards, a recent graduate from the School of Physiotherapy and Exercise Science has developed a prototype Web application to interact with the developed algorithm. Future work is planned to further develop the Web application to be accessible by researchers, students and teams with additional functionality to aid in conducting systematic reviews.