DIEM Analytics

DIEManalytics are specialists in applying AI to business needs, specifically within the defence sector. As part of the UK government’s Defence and Security Accelerator scheme, they wanted to progress their DUChESS (Dynamically Updated Chatbot for Extracting Summaries and System Understanding) chatbot.

The initial phases researched aggregating multiple interviews and improving end-user outputs using a Neuro-Linguistic Programming (NLP) engine. The next phase would look into making DUChESS a fully-networked application, partnering with MyOxygen/ to develop the front- and back-end for the web application.

• Back-end framework selection – DIEManalytics already had a working NLP engine prototype and database to store information. We would need to find a framework that could host the back-end that could communicate securely and efficiently with the database’s third-party secure hosting service.
• Front-end text-to-speech integration – the web application was designed to be run on multiple Operating Systems. This meant each OS would need its own implementation of text-to-speech functionality, further limiting framework selection.

Optimal user experience – system administrators would need clear and easy analysis of the data returned from the database. MyOxygen/ planned a typically user data and feedback-centric method of determining the optimal User Interface.

• Discovery Workshops
• MVP/Prototype
• Mobile Web App
• React Frontend
• Integration Azure
• AI

DUChESS now has a front-end that is fully integrated with the back-end, allowing users to carry out an interview quickly and efficiently.

Following framework and OS-specific text-to-speech functionality research, a solution was created that optimised the UI for system administrators. The framework selected for DUChESS integrates smoothly with both the third-party secure database and DIEManalytics’ NLP engine.

System administrators are now also able to analyse patterns and view aggregated words and phrases using a thematic map – indicated by user research to be the most accessible and effective way of viewing the data.