AI for Communication Symbols Categorisation

As part of our AI and Inclusion project, AI techniques were developed to automatically categorise communication symbols, enabling users to find more relevant symbols faster.

globalsymbols.com

Investigators

Professor Mike Wald

Principal Investigator

E.A. Draffan

Dr Russell Newman

Dr Chaohai Ding

Global Symbols is a repository of free-to-use communication symbols that allows any creator of Alternative and Augmentative Communication (AAC) symbols to publish their content with ease. Given the number of symbols available, it can be time-consuming for users to find relevant symbols for their purposes.

We developed AI techniques to automatically categorise and group symbols within the Global Symbols repository. Our technique uses metadata already known about the symbol (such as the label), a network of rich semantic data, and AI-based image recognition.

Most AI image recognition focuses on photographic images, which doesn't always transfer well to graphical symbols. We developed and refined processes that allow image recognition to be applied to graphics as well as photos.

This work will make it easier for AAC symbol users to find relevant symbols, either by working from a starting symbol, a subject, or an area of conversation.

Project originally funded by

The Alan Turing Institute is the UK's national institute for data science and artificial intelligence, with headquarters at the British Library.