How the Alt Text Gets Made: What Roles and Processes of Alt Text Creation Can Teach Us AboutInclusive Imagery

Abstract

Many studies within Accessible Computing have investigated image accessibility, from what should be included in alternative text (alt text), to possible automated, human-in-the-loop, or crowdsourced approaches to alt text generation. However, the processes through which practitioners make alt text in situ have rarely been discussed. Through interviews with three artists and three accessibility practitioners working with Google, as well as 25 end users, we identify four processes of alt text creation used by this company—The User-Evaluation Process, The Lone Writer Process, The Team Write-A-Thon Process, and The Artist-Writer Process—and unpack their potential strengths and weaknesses as they relate to access and inclusive imagery. We conclude with a discussion of what alt text researchers and industry professionals can learn from considering alt text in situ , including opportunities to support user feedback, cross-contributor consistency, and organizational or technical changes to production processes.

Using A Neural Network to Generate Images When Teaching Students to Develop an Alternative Text

“…Для автоматизации разработки альтернативных текстовых описаний к изображениям используются методы: программного извлечения подходящих текстов из пула текстовых описаний; компьютерного зрения; обработки естественного языка; глубокого обучения на основе нейронных сетей [11][12][13][14][15]. Вместе с тем перечисленные технологии имеют свои недостатки в отношении точности, полноты и корректности (в том числе этической) генерируемых описаний [11][12][13][14][15][16][17] по сравнению с текстовыми альтернативами, созданными человеком [14,15,17], которые также не всегда совершенны из-за недостаточной компетентности авторов [17,18].…”