Accessibility

Image descriptions are pieces of text accompanying an image which explain to readers the nature or contents of the image. Image descriptions are used by assistive technology to make the object accessible to people who have a visual disability and cannot read or see it. You might see it referred to as ‘alternative text’ or ‘alt text’.

We are adding image descriptions to improve the accessibility and discoverability of our content. Improving accessibility is rightly prioritised by governments around the globe, and adding image descriptions help us towards meeting the W3C’s Web Content Accessibility Guidelines (WCAG) 2.2 AA Standard and ensuring compliance with various accessibility-related legislation and standards, including Title II of the Americans with Disabilities Act (ADA).

For those journals typeset by Taylor & Francis typesetting suppliers, our suppliers will be using advanced AI technology to automatically produce image descriptions.

How AI produces image descriptions:

1. Image recognition: AI models, often powered by machine learning, use computer vision techniques to identify objects, scenes, and patterns within an image. For example, an AI system might recognize a “dog sitting on a grassy field” or “a bar chart showing sales growth.”

2. Natural language generation: Once the image is analysed, AI generates a textual description using natural language processing (NLP). This makes sure that the description is clear, concise, and human-readable.

3. Context awareness: Advanced AI systems can consider the surrounding content (e.g., captions, titles, or related text) to provide more contextually relevant descriptions. For instance, in a scientific journal, the AI might describe an image as “a graph showing the correlation between temperature and energy output.”

Our typesetting suppliers use advanced commercial AI services to generate image descriptions. A range of techniques are used to produce accurate and appropriate image descriptions, including:

1. Training on diverse datasets: AI models are trained on large, diverse datasets containing images and their corresponding descriptions. This helps the AI learn to recognize a wide range of objects and scenarios accurately.

2. Human review and editing: Our authors will have the opportunity to review and, if needed, edit image descriptions during the proofing process. For example, authors or editors may refine descriptions for technical diagrams or scientific images.

3. Incorporating domain-specific knowledge: AI systems can be customized to include knowledge specific to certain fields, such as medicine, engineering, or education. This makes sure that the descriptions are precise and relevant to the intended audience.

4. Avoiding bias and errors: Developers work to minimize biases in AI models by carefully curating training data and testing the system for accuracy across different contexts and demographics.

5. Feedback loops: AI systems improve over time by incorporating feedback from users, such as corrections or suggestions for better descriptions.

A long description is a detailed textual explanation of complex visual content, such as charts, graphs, diagrams, maps, or intricate illustrations, that cannot be adequately conveyed through brief alternative text (alt text). While alt text provides a concise summary of an image, a long description offers a more comprehensive explanation, making sure that users with visual impairments can fully understand the information and context presented in the visual element.

If an author provides their own image descriptions as part of the accepted manuscript we will use them instead of producing AI generated image descriptions.

We are happy to work with customers to process ad hoc requests for accessibility remediation. Requests can be submitted using this form.