[DRAFT: IN PROGRESS],
Beautiful official team photo:
Headline: Speech pattern metric calculation and database creation, enabled by patient story collection front-end and machine learning (NLP) on audio samples
Natural Language Processing (NLP) offers researchers the opportunity to analyze speech patterns of dementia patients and identify trends for dementia diagnosis and progression of the disease. However, we have only begun to scratch the surface of this, as researchers have limited access to large audio datasets and rarely have the technical know-how to use NLP to analyze audio. MemoryLog solves both of these problems.
MemoryLog uses a simple front-end to collect audio samples from dementia patients, and a combination of IBM Watson and prototyped audio analytics to create meaningful metrics for researchers and data scientists. We have also created a front end for researchers and physicians to easily visualize the data collected.
In the future, we hope to use virtual assistants (e.g. Alexa, Siri) to eliminate the patient front-end, increase the number and quality of metrics, and develop an interface for families to access and sort the contributed stories.
Python script with IBM Watson APIs and original code to perform NLP on voice recordings, calculate metrics, and output a CSV file
Tableau for dashboard with data visualization
Team (from left to right in photo):
Vithushan “Vitu” Jeyakumaran - Business, Mobile UI designer
Anand Ganeshalingam - Back-end developer
Bicheng “Dmitri” Liu - Back-end developer
Joshua Kanakaratnam - Front-end developer, data architecture
Jermain Joseph - Front-end developer, designer
Demo screenshots / photos:
The full story: