PharmaBot for Sanofi
Sanofi is a french drug company that invited us to build a chatbot to educate users about their new drugs. RIKAI Labs joined their accelerator program in Shanghai to apply our technology solution to this problem.
Chatbot with human fallback
The plan was to launch the bot first as an internal product, and use it to train the staff. In phase two launch it directly to doctor customers.
The format of questions was very formulaic with small differences only in technical product names. So I had to implement a new matching algorithm that combined using a classifier for the 'shape' and intent of the sentence with data from a customer specific knowledge graph.
As we improved the algorithm and trained the bot over multiple releases the matching rate improved. We check both false negatives and false positives.
Knowledge Graph and Natural Language Generation
We built a structured model of the questions specific to Pharmacologist, Based on this we could anticipate questions for many different products, and generate the questions and answers automatically.
End User Flow
The final bot combined chat knowledge with final product specs delivered as PDFs.
Completed Accelerator Program
We completed Sanofi's accelerator program in Shanghai to deliver the project.
Read more and see a video on RIKAI Labs case studies page here: http://www.rik.ai/case/sanofi