Using AI Language Models to Identify the Risk of Dementia
We built a complex AI language pipeline to flag dementia risk indicators in the speech patterns of elderly Japanese individuals for Nabetomo (株式会社Nabe), a Japanese startup that assigns conversation partners to elderly individuals through secure video sessions.
What We Did
We included robust integrations in our process to ensure many aspects of detecting dementia risk and providing value to participants were considered.
Tasks
Built a complex AI language pipeline
Calculate dementia indicators using speech complexity and answer cohesion
Build speech profiles for each elderly individual and trended their speech outcomes month-over-month
Visualized themes from elderly participants to share important moments from their lives with their family
3 Key Insights
Complex AI Pipeline
We knew that creating a solution that would meet Nabetomo’s needs would require a carefully established process.
Our AI pipeline:
-
Machine transcribed conversational speech patterns from almost a year’s worth of recorded video calls in Japanese.
-
Cleaned the data and separated it into meaningful sections for analysis.
-
Uses both the audio and transcriptions to investigate speech patterns, sentiment, and the topics discussed.
-
Processes and consolidates the data into benchmark measures and visualizes it in a secure web dashboard.
We worked closely with Nabetomo to ensure this process produced valuable insights.
AI Speech Analysis
The analysis is rigorous and captures the following features:
-
Complexity of words used
-
Question and answer cohesion
-
Percentage of conversation sharing
-
Speaker confidence and sentiment
-
Topics discussed
As an added value, this process captures discussions related to key memories, allowing participants to share them with their family (e.g., happy memory of a childhood birthday, experiences during a recent vacation, and tragic moments from living through World War II).
Benchmarking and Monitoring
These features were analyzed to produce a conversational score that can point towards the risk of dementia. This score was trended over time so participants and their family could quickly respond to changes in their health. Our models also flag security risks such as inadvertent discussions about personal banking. This helps keep participant’s data secure and protect their privacy.
Participating in this process contributes to encouraging early diagnosis of dementia and promoting social interaction, which can help prevent and delay the onset of dementia.