Our research for Digital Catapult focused on two specific areas of Natural Language Processing: abstractive text summarisation and extractive text generalisation. Following extensive research and testing, we discovered algorithms that would be most efficient at providing an insight into the contents of a passage, whilst also adapting existing methodologies that aim to automatically remove sensitive information about individuals and companies. BLEU (BiLingual Evaluation Understudy), METEOR (Metric for Evaluation of Translation with Explicit ORdering), SpaCY and numerous other metrics were utilised in the process of developing an algorithm that would be most effective at imitating human speech and limiting access to confidential information without distorting original meaning.

During our collaboration with Envitia Ltd on a mutual project for the UK Environment Agency, we were responsible for creating a system that would automatically recognise the altering water levels across the UK and warn the population of potential floods up to two weeks in advance. Through the power of Machine Learning, AI and Big Data analysis, a model was created to recognise the base water level via CCTV footage to subsequently estimate the probability and likely severity of a flood.

Whilst working with Methods Analytics, our primary role was to ensure that quality is embedded into the production process, and that the final product satisfies all key requirements. Therefore, process maps were integrated into the QA system, to enhance productivity and minimise losses. Furthermore, training was provided for all members of staff involved in the production cycle on how to manage projects and write code in the most efficient manner to prevent a duplication of efforts. Consequently, the long-run supportability and versatility of the code improved, with a library of standard functions created to increase the speed of development.