Time
1.5 months
Location
USA
Sector
Healthcare
Description
The client, an American healthcare company, had the goal of building a classifier to differentiate medical publications related to different types of diseases by analyzing abstracts.
Solution
To accomplish this task, the project utilized PubMed as a source of medical articles. PubMed is a comprehensive archive of biomedical and life sciences literature, with over 25 million references to journal articles. The solution involved the development of a fully-featured application that enabled the client to experiment with various classification models to achieve optimal results. Given the substantial volume of data, computations were conducted in parallel across a network of powerful AWS servers. To further optimize the process, several multiprocessing approaches were implemented.
Results
The project demonstrated the effectiveness of Natural Language Processing (NLP) in enhancing business processes. As the volume of scientific publications continues to grow, the client can now utilize machine learning to streamline searches and automatically identify medical publications related to specific topics. This significantly reduces the cost of information processing and search, resulting in cost savings of $72,000 at the initial stage of implementation.
Technologies
AWS, NLP tools (protected by NDA)