TORONTO, November 30, 2018
Two of our research papers were accepted and will be presented at NeurIPS (the conference previously known as NIPS) next week in Montreal.
Semi-supervised Classification by Reaching Consensus Among Modalities
(Zining Zhu, Jekaterina Novikova and Frank Rudzicz)
Deep learning has demonstrated abilities to learn complex structures, but they can be restricted by available data. We present Transductive Consensus Networks, suitable for semi-supervised learning, where different modalities of input are compressed into latent representations. TCNs outperform, or align with, the best benchmark algorithms given 20 to 200 labeled samples on the Bank Marketing and the DementiaBank datasets.
The Effect of Heterogeneous Data for Alzheimer’s Disease Detection from Speech
(Aparna Balagopalan, Jekaterina Novikova, Frank Rudzicz and Marzyeh Ghassemi)
Speech datasets for identifying AD are generally restricted to participants performing a single task (such as describing a picture) and are difficult to collect. Can we use easily available healthy data? We demonstrate that healthy data from multiple speech-based tasks helps improve AD detection by up to 9%. We also show the need for heterogeneous data for building models capable of AD detection across multiple tasks.
Winterlight Labs is commercializing a proprietary language-based diagnostic system that analyzes natural speech to detect and monitor dementia, Alzheimer’s, aphasia, and various other cognitive conditions. Winterlight's scalable platform uses short recorded speech samples to analyze hundreds of linguistic cues, and can detect dementia and other conditions with a high level of accuracy-. This is a major improvement over current pencil-and-paper tests which are time-consuming, costly, and difficult to administer. The platform has applications in drug trials, long-term and primary care, and speech-language pathology.
Since its founding, Winterlight Labs has seen strong interest from pharmaceutical companies and other potential partners who view the technology as a major improvement over current methods of detection or screening. The company has gained support from the pan-Canadian AGE-WELL Network of Centres of Excellence (NCE), the Ontario Brain Institute, Ontario Centres of Excellence, and the University of Toronto Banting and Best Centre for Innovation and Entrepreneurship. For more information, visit www.winterlightlabs.com.
Liam Kaufman, CEO
liam [at] winterlightlabs.com