Demonstration venue · read-only. Every page can be browsed; the buttons that would change it are switched off. Create an account to run TaxoReview on your own data. A simulation from public publication records: the people listed were not invited and did not agree to review.
? Tutorial for this page

Rosa Català

dblp:384/0511 · DBLP profile ↗
← Back
1ranked-venue papers
0as first author
1since 2021 · last 2024
0009-0003-0910-1190ORCID · reported

Domains — the database's venue-derived domains; a paper can count in several

Artificial intelligence and machine learning · 1 · 1 since 2021Databases, data management, data science and information retrieval · 1 · 1 since 2021

Expertise — from the expertise taxonomy: the topics of the expert's papers under the CCF categories. A weight counts papers with recency: 1 for a paper about the topic, 0.3 when the topic is its context, halved every five years.

Human-computer interaction and pervasive computing
1 paper
Accessibility and assistive technology · 100%

Topics — the 2 heaviest of 2, each with the papers that count most for it

TopicWeightPapersLastEvidence papers
Accessibility and assistive technology › web accessibility
social media accessibility
0.812024
Breaking Barriers: A Hands-On Tutorial on AI-Enabled Accessibility to Social Media Content · KDD 2024
Accessibility and assistive technology › assistive technology
AI for accessibility
0.212024
Breaking Barriers: A Hands-On Tutorial on AI-Enabled Accessibility to Social Media Content · KDD 2024

Methods — techniques the papers use, named apart from their topics

multimodal large language model · 0.8
YearPublicationVenuePosition
2024 Breaking Barriers: A Hands-On Tutorial on AI-Enabled Accessibility to Social Media Content
abstract
Reddit's mission is to bring community, belonging, and empowerment to everyone in the world. This hands-on tutorial explores the immense potential of Artificial Intelligence (AI) to improve accessibility to social media content for individuals with different disabilities, including hearing, visual, and cognitive impairments. We will design and implement a variety of AI-based approaches based on multimodal open-source Large Language Models (LLMs) to bridge the gap between research and real-world applications.
Julio Villena-Román, Rosa Català, Janine García, Concepción Polo, Yessika Labrador, Francisco del-Valle, Bhargav Ayyagari
KDD2