![](https://mathco.com/wp-content/uploads/2023/10/Retrieval-Augmented-Generation.webp)
Industry
Healthcare & Life Sciences
Region
US
Solution
RAG Processing Module
Context
Pharma companies allocate a significant portion of their revenue to market research activities. The information gathered from these activities is generally documented in PPT or PDF formats. Analysts from different departments manually review these documents to obtain vital insights. However, as the volume of information generated increases, manual methods are becoming increasingly difficult. As a result, many pharma companies are integrating LLMs or Q&A virtual assistants to optimize their market research process.
Problem statement
Our client, a renowned global pharmaceutical company, wanted to develop a market research Q&A virtual assistant to process unstructured data from different documents. But, one major obstacle they faced was with the RAG processing module required for this virtual assistant. The client’s processing module had some significant gaps, such as lower volume and quality of data extracted, semantic positioning, etc., which were hampering the functionality of the virtual assistant. We approached the client with a solution to enhance the performance of their RAG module, thereby optimizing their overall market research process.
Impact
- Increased the volume of extracted words by around 300%
- Maintained the semantic ordering and sequencing in about 70% of the slides
- Showcased around 90% accuracy on basic business queries when used with GPT-4