Peer-reviewed research
CARTGPT: Improving CART Captioning using Large Language Models
Liang-Yuan Wu et al., ASSETS 2024, work-in-progress paper, 2024
Read the originalRead on October 16, 2026
- What it is
- A University of Michigan paper that describes CART captioning, the live captioning done by trained stenographers, and how a language model can correct errors in CART and ASR captions. It reports that CART accuracy can decline in difficult conditions.
- How much weight it carries
- A peer-reviewed work-in-progress paper, so its results are preliminary. The comparison uses one noisy speech dataset and a small preliminary test with three DHH users.
Facts taken from this source
- The authors state that CART captioning accuracy can considerably decline, which affects how well deaf and hard of hearing people follow the content.
“the CART captioning accuracy can considerably decline, thereby affecting the comprehension of DHH people” (Abstract)
- The authors report that existing approaches for improving automatic speech recognition still fail to match trained CART captioners.
“these approaches still fail to perform at par with trained CART captioners” (Abstract)
Pages that rest on this source
- Human vs AI Transcription: Accuracy, Cost, SpeedGuide
- Stenography and Shorthand: Writing at 225 WPMGuide
- CART captioningGlossary
- Transcription: How to Turn Audio and Video Into TextTopic
- Speech to Text for Deaf and Hard of Hearing PeopleGuide
- Is automatic captioning accurate enough for deaf people?Question