Bridge’s approach to artificial intelligence (AI) tools has evolved rapidly, in keeping with global advances. In 2019, our innovations team began to explore whether tools such as natural language processing (NLP) can improve the efficiency and reach of literature studies, and published a summary of our initial findings.
With the advent of powerful new large language models (LLMs), Bridge has been running a systematic program to evaluate the performance of such models across key stages of literature studies – title/abstract (ti/ab) screening, full-text screening, extraction and reporting.
We have been training and testing these models with our gold-standard internal datasets, and validating them further using high-quality external datasets. Our latest findings are summarised in our white papers on title/abstract screening, full-text screening and prompt engineering.
Based on our results to date, we are offering clients AI-enabled approaches to certain aspects of literature studies - with a robust QC methodology in place to ensure the reliability of the final outputs.
At the same time, we are continuing to work on and improve the accuracy of our models and the value they offer. We will make our ongoing research available as each workstream completes.
To find out more about our AI research program and AI-enabled services, use the contact link below.
