A recent study has found a new application of Machine Learning (ML) and Natural Language Processing (NLP). These advancements can assess older adults’ subjective well-being by analyzing speech and offering more information than brief check-in conversations usually provide. The insights can be used to flag the need for interventions and give caregivers peace of mind.

“How are We Doing Today? Using Natural Speech Analysis to Assess Older Adults’ Subjective Well-Being” was published on Springer Link.

Why Speech Analysis Matters

Patients’ speech patterns and use of language can indicate a patient’s sentiment and even early signs of disease. The research drew from acoustic phonetics theory, which focuses on the sounds of speech, and prosody theory, which focuses on elements such as pitch, duration, and intensity of speech. Both show a person’s emotions, said the report.

The research aligned these theories with advancements in ML and NLP, which offer “promising potential in deriving subconscious information from the human voice.”

The research asked this question: How can older adult’s subjective well-being be assessed from natural speech?

The findings are important, because being aware of subjective well-being can enable older adults, especially those who live alone, to stay independent for longer. The study yielded promising new ways to better understand how older adults are doing.

About the Research

The research accepts that subjective well-being strongly influences physical and mental health, resilience, and other essential aspects of human life. The research uses two quality of life instruments developed by the World Health Organization (WHO), both of which measure subjective well-being.

The research also accepts that the human voice expresses both conscious and subconscious information like feelings and emotions. It applies NLP to assess feelings and emotions from natural speech, noting that this is a comparatively new area of research.

To create a data set, the researchers recruited 32 older adults between ages 60 and 88 to complete the two WHO quality of life questionnaires and participate in an interview. All participants were able to live independently and had no mental or cognitive challenges. Twenty female and 12 males participated.

The ML model predicts and gives numerical estimates of subjective well-being from natural language, calibrated to the WHO quality of life instruments.

Findings

The model was found to help detect unexpected shifts in subjective well-being, give precise numerical estimates of quality of life, and excel in quantitative evaluations. The research “confirms the subconscious transmission of an individual’s subjective well-being via natural language.”

Reimagining Digital Health Tools

The paper notes some traditional challenges in determining subjective well-being: That open-ended questions are more effective than the closed questions that are typically used, that asking repeated questions places a burden on older adults, and that self-reported data may not be reliable.

The ML model tested does not ask questions. As such, it is “a novel step toward reimagining digital health-based tools,” says the report. This model can monitor for as long as the older adult speaks and, because it creates objective measures, it may be more reliable than self-reported assessments. Daily monitoring of subjective well-being could help caregivers better identify any changes that warrant extra care.

The ML model could be calibrated to specific user groups, such as those who live active social lives versus those who live in isolation, said the report.

In addition, the ML model could help with patient monitoring around chronic disease management, new medication, and patients’ physical and mental healing and health.

To learn more, read the full study report.

Speech Analysis Algorithm Can Offer Screening for Cognitive Impairment

In another recent study, a machine learning tool performing speech analysis successfully identified cognitive impairment and dementia in older adults. The automated speech analysis algorithm could be used to screen older adults for cognitive impairment and aid in early detection.

The study, published in Frontiers in Neurology, was performed with people who speak Spanish. The tool distinguished between people who are cognitively normal and those who have mild cognitive impairment and dementia at an overall accuracy of 88%. An article in McKnight’s Senior Living, “New AI tool may help detect dementia’s earliest signs,” offers more detail.