The BBC provides a news service delivering advice and information about new developments in the health arena. But with some of the audience more likely to consume content on social media, rather than news pages, the BBC developed an experiment to generate a graphical health ‘comic’ that could be posted to Instagram.

To achieve this, it was necessary to create a taxonomy of health concepts and their synonyms, mapping them to a set of related images. These images were further classified into types and given properties, so that they could be combined to give a composite image that represented the text in a meaningful way. They were then able to extract the entities from the text and match them to images, combining them to the panels of a comic that could be posted to Instagram. Jo describes the entire taxonomy creation process, outlining how the original story moved through the stages to produce the end comic.

Health Education England (HEE) exists to ensure the NHS workforce has the right numbers, skills, values, and behaviours. The organisation handles a myriad of content – from evidence to inform policy decisions and development work, to producing and publishing e-learning modules - which is available to all healthcare staff in England. Ordinarily, this content includes a variety of topics, is rapidly developing, and is accessed by a variety of user groups.  Development of new taxonomies had already begun as part of a long term project to improve the range of content and the user experience.

The Coronavirus pandemic has accelerated the need for e-learning content and materials in particular, which must be rapidly produced and curated in a range of modules to help all levels of healthcare staff inform and improve their skills and confidence. The taxonomy project needed to adapt to take this into account and ensure fast and easy access to this new content.

Emily and Richard discuss the approach taken in selecting and adapting existing taxonomies for this purpose, consulting with users, and ensuring a balance between ease of access and discoverability of content with robust indexing systems, while ensuring the flexibility to rapidly adapt to new content and use.

Vagueness is a common human knowledge and language phenomenon, demonstrated by terms and concepts with blurred boundaries – for example, ‘tall’ and ‘expert’ - whose expansions are difficult to determine precisely, but then building taxonomies, ontologies and other types of semantic data models, modelers and domain experts often need to use such vague concepts. If this is not done properly, then these vague concepts negatively influence understanding by users, limiting their value as a source of knowledge.

Panos presents a set of practical strategies and techniques for tackling vagueness in semantic data modelling to create models that are semantically more accurate and interoperable. Key take-aways include: