Natural language algorithms to write weather forecasts

Computer-generated weather updates which mimic prose as if written by a human operator are being tested by computer scientists at Heriot-Watt University and University College London.

Met Office data will be used and the BBC will provide its editorial expertise to help develop new algorithms for automatically generating weather reports. If the project is successful, a prototype system will be tested out in generating local weather reports on a suitable part of the BBC’s website.

Dr Verena Rieser of Heriot-Watt University’s School of Mathematical and Computer Science, and her co-investigators Dr Andreas Vlachos and Dr Sebastian Riedel from the Computer Science department at University College London (UCL), have been awarded more than £560,000 for the two-year project to develop a technology that will translate vast and complex data, such as weather information, and turn it into an easily read, ‘naturally’ written communication. The software takes weather statistics and information and then weaves them into a written narrative.

The project is titled Domain-Independent Language Generation (DILiGENt) and the funding comes from the Engineering and Physical Sciences Research Council (EPSRC) for Dr Rieser and her team of co-investigators from University College London and Heriot-Watt University. Dr Rieser said, “The process is known as ‘natural language generation’ (NLG) and creates a more flexible output for interactive interfaces. It will overcome one of the main problems for intelligent personal assistants such as Apple’s Siri or Google Voice which is the often repetitive responses to questions or directions.

“Machine learning for Natural Language Processing (NLP) is what DILiGENt is all about. Large companies such as Microsoft, Amazon and Yahoo are making substantial new investments in this field and there are start-up opportunities emerging. The UK is a world leader in this kind of research but we are under-represented in using machine learning approaches for NLG. This project will help to strengthen this research strand in the UK.”