Automated German Alt Text Generation for News Charts

Alessia Vannini, Julia Locher, Marianne Santaholma, Claudia Amsler, Oesch Jonas, Arzu Coltekin

Track: Scientific Track

Keywords: Chart alt text generation, alt text, MLLM, gold standard corpus, LLM as judge

TL;DR: This paper investigates whether MLLMs can automatically generate accessible, gold-standard-quality German chart alt texts for visually impaired users, evaluated both quantitatively and qualitatively.

Abstract: We investigate whether a Multimodal Large Language Model (MLLM) can automatically generate well-formed German chart alt texts that meet the requirements of visually impaired persons while following the accessibility guidelines for chart alt texts, and match the quality of manually authored gold-standard alt texts. Focusing on bar, line, and stacked bar charts from a German-language newspaper (the Neue Zürcher Zeitung), we define an alt text structure, construct a gold-standard corpus, and evaluate MLLM-generated chart alt texts both quantitatively on semantic similarity and qualitatively with visually impaired persons as participants in terms of clarity, conciseness, meaningfulness, and output consistency.

Submission Number: 51