Analysis of spontaneous speech in Parkinson’s disease using natural language processing
Highlights • Spontaneous speech of patients with Parkinson’s disease (PD) was analyzed. • PD patients spoke fewer morphemes in a sentence than healthy controls. • PD patients had a higher rate of verbs and a lower rate of nouns than controls. • Discrimination accuracy rates using identified language items were greater than 80%. |
Patients with Parkinson’s disease (PD) face a variety of speech-related problems, including dysarthria and language disorders. To elucidate the pathophysiological mechanisms of linguistic impairment in PD, we compared the pronunciation of patients and that of healthy controls (HC) using automated morphological analysis tools.
Methods
We enrolled 53 Parkinson’s disease (PD) patients with normal cognitive function and 53 HC, and assessed their spontaneous speech using natural language processing. Machine learning algorithms were used to identify characteristics of spontaneous conversation in each group. Thirty-seven features focusing on part of speech and syntactic complexity were used in this analysis. A support vector machine (SVM) model was trained with ten-fold cross-validation.
Results
Patients with Parkinson’s disease (PD) were found to speak fewer morphemes in a sentence than the group of healthy controls (HC). Compared with HC, PD patients’ speech had a higher rate of verbs, case particles (dispersion), and verb utterances, and a lower rate of common noun utterances, proper noun utterances, and filler utterances. . Using these conversational changes, the respective discrimination rates for PD or HC were over 80%.
Conclusions
Our results demonstrate the potential of natural language processing for linguistic analysis and diagnosis of PD.
Comments
Using artificial intelligence (AI) to process natural language, a research group evaluated speech characteristics among patients with Parkinson’s disease (PD). AI’s analysis of their data determined that these patients spoke using more verbs and fewer nouns and fillers. The study was led by Professor Masahisa Katsuno and Dr. Katsunori Yokoi of Nagoya University School of Medicine, in collaboration with Aichi Prefectural University and Toyohashi University of Technology. They published their results in the journal Parkinsonism & Related Disorders .
Natural language processing (NLP) technology is a branch of AI that focuses on enabling computers to understand and interpret large amounts of human language data using statistical models to identify patterns. Since PD patients experience a variety of speech-related problems, including impaired speech production and language use, the group used NLP to analyze differences in the patients’ speech patterns based on 37 characteristics using texts prepared from free conversations.
The analysis revealed that PD patients used fewer common nouns, proper nouns, and fillers per sentence. On the other hand, they spoke using a higher percentage of verbs and case particle variance (an important feature of the Japanese language) per sentence.
According to Yokoi, “When I asked them to talk about their day in the morning, a PD patient might say something like the following, for example: ’I woke up at 4:50 am. I thought it was a little early, but I got up. It took me about half an hour to go to the bathroom, so I got washed and dressed around 5:30 am. My husband prepared breakfast. I had breakfast after 6 am. Then I brushed my teeth and got ready to go out.’”
Yokoi continued: “Whereas someone in the healthy control group might say something like this: ’Well, in the morning, I woke up at six o’clock, got dressed, and, yes, I washed my face.’ Then, I fed my cat and dog. My daughter prepared a meal, but I told her she couldn’t eat and I, umm, drank some water.’”
“While these are examples we created of conversations that reflect the characteristics of people with PD and healthy people, what you should see is that the overall duration is similar,” Yokoi explained. “However, PD patients speak shorter sentences than people in the control group, resulting in more verbs in the machine learning analysis. The healthy control also uses more fillers, such as ’good’ or ’umm’ , to connect sentences.”
The most promising aspect of this research is that the team conducted the experiment in patients who did not yet show the characteristic cognitive impairment seen in PD. Therefore, their findings offer a potential means of early detection to distinguish PD patients.
“Our results suggest that, even in the absence of cognitive impairment, the conversations of PD patients differed from those of healthy subjects,” concludes Professor Katsuno, director of the study. “When we attempted to identify PD patients or healthy controls based on these changes in conversation, we were able to identify PD patients with over 80% accuracy. “This result suggests the possibility of language analysis using natural language processing to diagnose PD.”
