Voody is a cognitive search and extraction platform for clinical research documents. The user will be able to create clinical trial search spaces into which, trial related documents like protocol, SAP, TMF contents and external research articles can be uploaded. Voody also has a provision to crawl content from an external website or connect to document management systems like Dropbox, Sharepoint and Documentum. The user will then be able to ask questions about the trial using a natural language interface and Voody will provide exact answers for your questions from unstructured text, tables and images embedded in the documents.
Voody used a computer vision based machine learning model to understand the documents and create a searchable index. It also comes with an inbuilt crawler to help users to assimilate contents from a website.
Our search is semantic in nature as opposed to a token based search. So even if the questions asked does not have the exact token, Voody will identify the documents of interest and highlight the portions of the document which contains the answer.
We have developed a robust machine comprehension module which will extract the exact answer for the user queries. This module is built on the google Bert vectors and further refined to make it fast and scalable.
The search space module of Voody will allow the users to structure their documents for each trial and also manage the user rights. The users will be able to collaborate on a single search space as a contributor or an analyst.