Case Study: Leading Pharmaceutical Company achieves highly accurate, 10× scalable adverse-event detection with TransPerfect's DataForce

Using Natural Language Processing for Adverse Event Detection

Leading Pharmaceutical Company engaged TransPerfect to support development of an NLP algorithm for adverse event (AE) detection in social media. The client needed more than 10,000 social posts categorized into predefined named entities to improve patient experience and reduce false negatives, with very high accuracy expectations given the sensitive medical context; TransPerfect’s DataForce platform and services were selected to provide annotation and human-in-the-loop support.

TransPerfect deployed DataForce’s global community to handpick pharmacovigilance-experienced reviewers, used double-blind annotation, Cohen’s Kappa quality metrics, and a reconciliation workflow, and iteratively updated guidelines as new edge cases appeared. This human-in-the-loop approach allowed the team to scale quickly—reviewing thousands of posts in a few days and increasing scalability by about 10×—and the client reported the TransPerfect solution was more scalable and efficient than traditional manual categorization.