Providing descriptive captions for images. How important is this for authors and publishers?

Posted on : November 28th 2022

Descriptive Caption for Images – Introduction

Each organization's decision to manage its image description workflows is influenced by a variety of factors. It is however a universal truth that you will be closer to achieving fully accessible content if you can make image description an integral part of content creation. In-text descriptions and captions that accompany figures, diagrams, and pictures are required but are frequently brief and insufficient.

WebAIM, a non-profit group affiliated with Utah State University that focuses on web accessibility, analyzed one million home pages in February 2021 and discovered that 60.6% lacked alt text. In 2019, a study conducted by Carnegie Mellon on 1.09 million tweets containing images revealed that barely 0.1% of these tweets contained alt text.

Researchers of all kinds, now more than ever, have the capability and responsibility to enrich scholarly outputs with the kind of information that only they, as authors, can offer, such as detailed descriptions of images that they have used. Those who rely on screen readers or other text-to-speech apps will not be able to perceive images properly if they are not accompanied by detailed and precise descriptions. Readers who lack the required software or technology to view the original copies may likewise be unable to access these images.

Image descriptions are crucial, and it is important to get them right

The responsibility of ensuring that every image, equation, and other visual asset has an accompanying detailed description lies with the publishers. Employing inclusive design principles allows for the continuous analysis of varied reader experiences in the development of content platforms and services. Readers with restricted vision and those without a device or software that can view images in scholarly publications will lose a significant amount of information provided by scholarly authors and editors if image descriptions are not included in the underlying metadata.

Authors are in the greatest position to know what an image in a scholarly publication is intended to express. Acquiring draft image descriptions from authors at the manuscript stage itself enables editors and service providers to improve those descriptions for accessibility in the context of enhancing and rendering the content in which the images feature.

Authors, on the other hand, may be unaware of the accessibility criteria underlying image description, as well as the technological requirements of this additional content. When it comes to writing descriptions that are significant and pertinent, some publishers offer training to their authors, and in some cases, this criterion is already included in the authors' contracts. This might not be feasible for everyone, in which case third-party vendors with relevant experience or internal editorial or production employees may be called upon. The key is to explicitly incorporate this step of the production process into the workflow so that it automatically becomes a part of the Digital ebook/ Courseware development cycle.

Using Artificial Intelligence

It makes perfect sense to use cutting-edge technologies, like artificial intelligence, to generate image descriptions. Microsoft, Google, and Facebook all have tools that can evaluate an image and suggest an alternate text description, and as technology continues to evolve, visual descriptions should get more precise.

The intricacy of the images in the publishing business is high when compared to other industries. Hence, the standard solutions that are available are insufficient on their own. The majority of graphic content and images in complex layout books such as schoolbooks, and scientific, academic, and professional publications are not photographs, but rather drawings or illustrations including infographics, graphics, complicated imagery, diagrams, scientific schemas, etc. A new generation of algorithms will be necessary for these kinds of images.

Despite the fact that artificial intelligence and automation have enabled alt text to be generated on a much larger scale, visually impaired academics around the world believe that the descriptions are typically of poor quality. Some advocates for disability rights assert that AI-generated alt text is improving, but still frequently misses the context.

Using AI technology to enhance the accessibility of images does not seem to be a practical solution for the publishing industry. Embedding description authoring into contracts and workflows seem to be the only practical approach at this time. The potential of this technology is however evident, and with the help of better algorithms, larger data sets, and maybe even analyzing the image in the context of any text around it, the precision and quality of automatically generated images have the potential to improve significantly.

In Part 2, we will discuss methods for making images accessible to everyone.