Summary: OpenAI’s ChatGPT program can identify spontaneous speech cues that are 80% accurate in predicting the early stages of dementia.
Source: Drexel University
The artificial intelligence algorithms behind the ChatGPT chatbot program, which has drawn attention to its ability to generate human-like text responses to some of the most creative queries, may one day help doctors detect the early Alzheimer’s disease.
Research from Drexel University’s School of Biomedical Engineering, Science and Health Systems recently demonstrated that OpenAI’s GPT-3 program can identify spontaneous speech cues that are 80% accurate in predicting early stages of speech. Madness.
Reported in the newspaper Digital Health PLOSThe Drexel study is the latest in a series of efforts to show the effectiveness of natural language processing programs for the early prediction of Alzheimer’s disease, building on current research suggesting that the disorders language may be an early indicator of neurodegenerative disorders.
Find a warning sign
Current practice for diagnosing Alzheimer’s disease typically involves a review of medical history and a long series of physical and neurological assessments and tests. Although there is still no cure for the disease, detecting it early can give patients more treatment options and support. Since language impairment is a symptom in 60-80% of dementia patients, researchers have focused on programs that can detect subtle cues, such as hesitation, grammar and pronunciation errors and forgetting the meaning of words, as a quick test that could indicate whether or not a patient should undergo a full examination.
“We know from ongoing research that the cognitive effects of Alzheimer’s disease can manifest in language production,” said Hualou Liang, Ph.D., professor at Drexel’s School of Biomedical Engineering, Science and Health Systems and research co-author.
“The most commonly used tests for the early detection of Alzheimer’s disease examine acoustic characteristics, such as pause, articulation and voice quality, in addition to cognition tests. But we believe that improving natural language processing programs offers another avenue to support the early identification of Alzheimer’s disease.
A program that listens and learns
GPT-3, officially the third generation of OpenAI’s General Pretrained Transformer (GPT), uses a deep learning algorithm, trained by processing large swathes of information from the internet, with a particular focus on how the words are used and the way the language is constructed. This training allows him to produce a human response to any task involving language, from answering simple questions to writing poems or essays.
GPT-3 is particularly good for “learning without data”, meaning it can answer questions that would normally require external knowledge that has not been provided. For example, asking the program to write “Cliff’s Notes” of a text would normally require an explanation, i.e. a summary. But GPT-3 has had enough training to understand the reference and adapt to produce the expected response.
“GPT3’s systemic approach to language analysis and production makes it a promising candidate for identifying subtle features of speech that may predict the onset of dementia,” said Felix Agbavor, PhD researcher at the ‘School and main author of the article.
“Training GPT-3 with a massive set of interview data, some of it involving patients with Alzheimer’s disease, would provide it with the information it needs to extract speech patterns that could then be applied to identify markers in future patients.”
Searching for voice signals
The researchers tested their theory by training the program with a set of transcriptions from part of a dataset of voice recordings compiled specifically for the purpose of testing the ability of natural language processing programs to predict the dementia. The program captured significant features of word use, sentence structure and text meaning to produce what the researchers call “integration” – a characteristic pattern of Alzheimer’s disease speech.
They then used the integration to repurpose the program, turning it into an Alzheimer’s screening machine. To test it, they asked the program to look at dozens of transcripts from the dataset and decide whether or not each had been produced by someone who was developing Alzheimer’s disease.
Running two of the best natural language processing programs at the same rate, the group found that GPT-3 performed better than both, in terms of accurately identifying Alzheimer’s examples, identifying non-Alzheimer’s examples, and with less missed cases than both programs.
A second test used textual analysis of GPT-3 to predict the score of various patients from the data set on a common test to predict the severity of dementia, called the Mini-Mental State Exam (MMSE) .
The team then compared the accuracy of the GPT-3 prediction to that of an analysis using only the acoustic characteristics of the recordings, such as pauses, voice loudness and slurring, to predict the MMSE score. The GPT-3 was found to be almost 20% more accurate in predicting patients’ MMSE scores.
“Our results demonstrate that text integration, generated by GPT-3, can be reliably used not only to detect individuals with Alzheimer’s disease from healthy controls, but also to infer test scores. subject cognitive data, both based solely on voice data,” they wrote. .
“We further show that text embedding outperforms the conventional acoustic feature-based approach and even performs competitively with fine-tuned models. These results, taken together, suggest that GPT-3-based text embedding is a promising approach for the assessment of AD and has the potential to improve the early diagnosis of dementia.
Continue the search
To take advantage of these promising results, the researchers plan to develop a web application that could be used at home or in a doctor’s office as a screening tool.
“Our proof of concept shows that it could be a simple, accessible, and sensitive enough tool for community testing,” Liang said. “This could be very useful for early detection and risk assessment before a clinical diagnosis.”
About this AI research news
Original research: Open access.
“Predicting dementia from spontaneous speech using large language modelsby Felix Agbavor et al. Digital Health PLOS
Predicting dementia from spontaneous speech using large language models
Language disorders are an important biomarker of neurodegenerative disorders such as Alzheimer’s disease (AD). Artificial intelligence (AI), in particular natural language processing (NLP), has recently been increasingly used for the early prediction of AD through speech. Yet relatively few studies exist on the use of large language models, particularly GPT-3, to aid in the early diagnosis of dementia.
In this work, we show for the first time that GPT-3 can be used to predict dementia from spontaneous speech. Specifically, we exploit the vast semantic knowledge encoded in the GPT-3 model to generate text embedding, a vector representation of text transcribed from speech, which captures the semantic meaning of the input.
We demonstrate that text embedding can be reliably used to (1) distinguish individuals with AD from healthy controls and (2) infer the subject’s cognitive test score, both based solely on speech data.
We further show that text embedding significantly outperforms the conventional acoustic feature-based approach and even performs competitively with current fine-tuned models.
Together, our results suggest that GPT-3-based text integration is a viable approach for assessing AD directly from speech and has the potential to improve early diagnosis of dementia.