Harnessing AI for Pharma R&D

The Challenge

The client needed an automated system to extract English and Japanese publications across diverse databases using the same search strategies applied in English queries to identify KOLs without expert intervention.

We leveraged machine learning and natural language processing to build a unified search system, optimizing cross-language database searches. The solution seamlessly integrated with an existing search engine to extract Japanese articles from English queries, saving translation time. We used the following techniques to build the solution:

Word Embedding

Using the Universal Medical Language System, we created word embeddings for both English and Japanese. Word embeddings represent words as unique real-valued vectors, where similar words share the same values. This sharpens the computer’s understanding of word context and relationships in text.

Nearest Neighbor Algorithm

We implemented the K-Nearest Neighbor algorithm to fetch relevant results, facilitating English queries to retrieve equivalent Japanese and closely related terms.

By integrating cutting-edge natural language processing and machine learning methodologies, Mu Sigma engineered a powerful system capable of generating an adaptive and intelligent solution, organically enhancing the embedded search strategy.

This system distinguishes itself from conventional approaches by considering multiple variables beyond simple term matching, thereby delivering the most pertinent results to queries. Furthermore, it demonstrated its preparedness for seamless expansion to encompass a diverse range of languages beyond English and Japanese.