Automated, Context-Aware Alt Text Generation for Educational Documents Using Large Language Models
Open accessAbstract
This paper presents ALADDIN, a tool developed by [anonymized] to automate the generation of alternative text (alt text) for images in Microsoft Word and PowerPoint files, aimed at improving the accessibility of educational materials. The tool leverages Gemini Flash, a large language model (LLM), to generate context-aware and content-appropriate image descriptions by analyzing the surrounding text, slide content, existing metadata, and visual layout. Currently integrated into a user-friendly Google Colab notebook, ALADDIN has reduced alt text remediation time for large files of educational materials from weeks to hours. We describe the tool's implementation, evaluation, and future directions for broader accessibility enhancement.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
AI-generated educational content can become substantially more accessible when the generating tool is deliberately configured.
“when generative AI tools are given explicit, operationalized accessibility instructions, through carefully engineered prompts [15,16], structured remediation pipelines [17,18], or context-aware generation conditioned on the surrounding material [19], measured compliance improves substantially compared to unconfigured use of the same tool.”
From A Protocol for Evaluating the Accessibility of AI-Generated Educational Materials: Prompt Configuration, WCAG-Derived Criteria, and Content Overload (2026) · cites this paper as [19]
Section 2.3 (Context Extraction; page unavailable) describes configuring Gemini to use surrounding text, slide content, and other context when generating alt text. Section 4 (Evaluation; page unavailable) says ALADDIN’s descriptions were more closely related to the document topic than Microsoft’s, supporting a quality benefit. It does not report measured accessibility-compliance gains against unconfigured use of the same tool. (LLM, conf=0.87)
Experimental — accuracy not yet measuredThe improvement observed under Condition B extends to educational content specifically findings already reported for web code, PDF tagging, and alt-text generation individually.
“The improvement observed under Condition B (Section 4.1) would extend, to educational content specifically, findings already reported for web code [15,16], PDF tagging [18], and alt-text generation [17,19] individually.”
From A Protocol for Evaluating the Accessibility of AI-Generated Educational Materials: Prompt Configuration, WCAG-Derived Criteria, and Content Overload (2026) · cites this paper as [19]
When LLMs are used as a collaborative tool, they show promise in assisting authors to write alt text.
“Nevertheless, when LLMs are used as a collaborative tool, they show promise in assisting authors to write alt text [3, 71].”
From How University Disability Services Professionals Write Image Descriptions for HCI Figures Using Generative AI (2026) · cites this paper as [3]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Context-Aware Image Descriptions for Web Accessibility
Mohanbabu, A. G., & Pavel, A. (2024) · International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) · doi:10.1145/3663548.3675658 · Cited by 29 · see citing works
Flagged first, then by citations
Cited by
Sorted by citations
Raees, M., Iwamoto, Y., Papangelis, K., Heard, J., & Tigwell, G. W. (2026) · arXiv · doi:10.48550/arxiv.2602.08937 · Cited by 1 · see citing works
“However, DSO professionals, often tasked to meet regulatory compliance and accessibility needs, face challenges with writing alt text for figures in scientific publications outside their”
Amado-Salvatierra, H. R. (2026) · arXiv · doi:10.48550/arxiv.2608.00749
“…extraction-and-context pipeline [17], and a related system generates alt-text for images in educational documents by conditioning generation on surrounding context rather than the image alone, addressing exactly the “purpose blindness” limitation Cardia et al. documented in Section 2.4 [19].”
InfluentialALADDIN: Efficient, High-Quality, AI-Assisted, Human-in-the-Loop ALT Text Generator
Baglodi, D., Martincic, I., & Walker, B. N. (2026) · International Web for All Conference (W4A) · doi:10.1145/3800424.3800439
Self-citationAdvancing the Taxonomy for Writing Alternative Text Based on Image Type, Context, and Complexity
Martincic, I. (2026) · International Web for All Conference (W4A) · doi:10.1145/3800424.3800440
Turkarslan, A., Thakkar, V., Tekle, R., Chen*, T., Martinez, J. J., & Mankoff, J. C. (2026) · International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) · doi:10.1145/3797867.3829048
Song, H., Shin, M., Choi, J., Kim, Y., Jang, K., Suh, B., & Jung, H. (2026) · Proceedings of the 2026 28th International ACM SIGACCESS Conference on Computers and Accessibility · doi:10.1145/3797867.3828985
- Authors
- Baglodi, D., Martincic, B., Sinclair, N., & N. Walker, B.
- Affiliations
- Georgia Institute of Technology (US)
- Psychology, Georgia Institute of Technology, Atlanta, Georgia, USA
- Venue
- Proceedings of the 27th International ACM SIGACCESS Conference on Computers and Accessibility
- Publisher
- ACM
- Published
- 2025-10-22
- Pages
- 1-4
- OpenAlex
- W4415444813
- Cited by
- 4
- References
- 1
- Citations by year
- 2026: 4
- FWCI
- 4.90
- Type
- conference paper
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) (publishedVersion)
- Topics
- Social Sciences › Human Factors and Ergonomics › Digital Accessibility for Disabilities
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
- Physical Sciences › Computer Science › Information Systems › Artificial Intelligence in Education
- Keywords
- picture description, accessibility, teaching materials, Microsoft Word, PowerPoint, large language models, Gemini Flash, Aladdin
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
