Key outcomes Meet-up Bologna Children's Book Fair: image description and accessible publishing

Key findings, struggles

Workflows

Workflows regarding image description may vary a lot, depending on the publisher: some publishing houses rely on external providers, some work in a hybrid way (with editors and AI support), others are doing image description internally with or without a specific training. In almost all cases, human action is still needed. Editorial departments take care of revising the alternative descriptions. Challenges are present at different levels in any organization. It is important to offer training and awareness to publishers so that they do not perceive alternative description only as a burden but as something that grants equal access to resources.

One big question is who should write the alt text. It is complex to understand at which stage of the workflow this theme is addressed. In some countries like in the UK or US or Canada, some publishers are starting to roll out contracts for authors where they specifically say that whenever you submit a manuscript with images, you also need to provide alt texts. This implies that authors should have some expertise.

Sometimes, for images that carry a precise message or meaning, a person with specific knowledge or cultural context is needed to write the alt text. Writing a good image description not only requires having a general knowledge of how to do it but also applying it in a specific context. Sometimes editors take care of the image descriptions itself or do a quality check.

Another key element is the testing phase with end users. It is very important to test publication, but also to test image description with people that will have to rely on this image description. Organizing focus groups allows us to verify if the descriptions made are efficient. Moreover, it could be very useful to gain more insights into what is missing or needs further adjustments, or maybe if the text needs to be simplified.

Frequently asked questions by publishers

• Level of detail in description
• Exemptions
• Disproportionate burden
• Decorative images

Trade and educational

In schoolbooks, images are a big part of the educational program, it is thus necessary that the images are precisely described. On the contrary, in trade books there is no need for the image description to be exactly right as long as we create a concept, an idea that can help the person to understand.

But, especially in illustrated children’s books, images are crucial to allow children to understand the meaning of the story because usually pictures tell more of the story than the book itself does. Nevertheless, children’s books are excluded from the European Accessibility Act. Children’s books in digital are mostly made in fixed layout and they are more difficult to make accessible, it will need one more workflow.

Concerning decorative images for educational books, there is no need to cut them out. The decorative part is still appealing for people who can see it; thus, it is worth keeping it and simply tag images as decorative. Overall, one of the biggest misconceptions about accessibility is that you have to cut something out in order to have an accessible version. For example, regarding the theme of colors, many companies outside the publishing world think that an accessible publication needs to be in black and white, or that images need to be cut out, but it is not like that. That is also why specific training is needed. The advice is to just start to describe the images, progress over perfection.

The context

Understanding the context in which an image is embedded is very important for elaborating a good image description. The same image in different contexts can be described differently. The depth of the description also may vary according to the context.

For example: if a monument, like the Eiffel Tower, is presented in an art book it will be described in depth, but if you find it in a French language book the description will be different. It is a fine line, and it is very difficult because guidelines say nothing about that. Given an image we can have more than one good alt text.

Categories are crucial for publisher and for consistency of the user. Be consistent on how you give information is very much appreciated.

AI

How AI is coping with the context will be crucial in the future. When using AI, one of the important questions in terms of describing is: “what is the purpose of the image?”.

AI models can be used to generate images. So, if you have a description, you can give your description to the AI model, and the model can generate the image according to the description. In this way, you will have the alternative text and the long description, and in the image, if you are happy with the image that the model can generate for you.

AI can create alt texts. With GPT, for example, you can add some text, and the model can follow this text, these sources, this information. The point is starting to work with these models and follow their evolution. It is important to notice that, like the society who trained it, also the AI is subject to bias. AI also has limits, for example hallucinations, meaning conveying false information. Moreover, it is worth noticing that AI has difficulties mostly in generating alt texts for STEM. It could be a good starting point but the human check is still necessary.