Shane Settle
Papers
1
Total Citations
12
H-Index
1
About
Shane Settle is a researcher at the intersection of speech processing, multimodal learning, and artificial intelligence, with a focus on how machines can learn language from raw, untranscribed audio. His most-cited work, "Visually Grounded Learning of Keyword Prediction from Untranscribed Speech" (2017, 12 citations), addresses a fundamental challenge in AI: how to bridge the gap between spoken language and visual perception without relying on text. By leveraging paired audio and visual data—mimicking the way infants learn from their environment—Settle demonstrated that neural networks can predict keywords directly from speech signals, enabling tasks like image retrieval using spoken queries. This contribution is pivotal for developing more natural, human-like AI systems that learn from multimodal sensory input, with implications for robotics, assistive technologies, and low-resource language processing. Though his citation count is modest, the work’s conceptual novelty and interdisciplinary appeal have made it a touchstone for researchers exploring grounded language acquisition. Settle’s research continues to push the boundaries of unsupervised and visually grounded speech learning, offering a compelling blueprint for machines that understand language as humans do—through context, not transcription.
Research Focus
Key Achievements
Top Papers
- 1Visually Grounded Learning of Keyword Prediction from Untranscribed Speech12 citations · 2017