Automatically Identifying Twitter Users for Interventions to Support Dementia Family Caregivers: Annotated Data Set...

Area of Science:

• Computational linguistics
• Social media analysis
• Public health

Background:

• Over 6 million Americans have Alzheimer disease and related dementias, relying on informal caregivers.
• Existing interventions for family caregivers are often inaccessible.
• Family caregivers of people with dementia utilize Twitter to share experiences.

Purpose of the Study:

• To develop an annotated dataset for identifying Twitter users with family members suffering from dementia.
• To create benchmark classification models for this identification task.

Main Methods:

• Collected 10,733 tweets mentioning dementia keywords and familial relationships.
• Annotated tweets to identify users with family members suffering from dementia.
• Trained and evaluated machine learning models, including deep neural networks (BERT).

Main Results:

• A BERT-based deep neural network achieved an F1-score of 0.962 in identifying relevant tweets.
• The model identified 128,838 tweets from 74,290 users indicating a family member with dementia.
• This demonstrates a scalable approach to identifying this population on Twitter.

Conclusions:

• An annotated dataset and classification models can automatically identify Twitter users with family members suffering from dementia.
• This capability facilitates large-scale research into family caregiver experiences.
• It also enables direct targeting of interventions to support these caregivers.