Automatic Speech Recognition of Conversational Speech in Individuals With Disordered Speech.

PURPOSE: This study examines the effectiveness of automatic speech recognition (ASR) for individuals with speech disorders, addressing the gap in performance between read and conversational ASR. We analyze the factors influencing this disparity and the effect of speech mode-specific training on ASR accuracy. METHOD: Recordings of read and conversational speech from 27 individuals with various speech disorders were analyzed using both (a) one speaker-independent ASR system trained and optimized for typical speech and (b) multiple ASR models that were personalized to the speech of the participants with disordered speech. Word error rates were calculated for each speech model, read versus conversational, and subject. Linear mixed-effects models were used to assess the impact of speech mode and disorder severity on ASR accuracy. We investigated nine variables, classified as technical, linguistic, or speech impairment factors, for their potential influence on the performance gap. RESULTS: We found a significant performance gap between read and conversational speech in both personalized and unadapted ASR models. Speech impairment severity notably impacted recognition accuracy in unadapted models for both speech modes and in personalized models for read speech. Linguistic attributes of utterances were the most influential on accuracy, though atypical speech characteristics also played a role. Including conversational speech samples in model training notably improved recognition accuracy. CONCLUSIONS: We observed a significant performance gap in ASR accuracy between read and conversational speech for individuals with speech disorders. This gap was largely due to the linguistic complexity and unique characteristics of speech disorders in conversational speech. Training personalized ASR models using conversational speech significantly improved recognition accuracy, demonstrating the importance of domain-specific training and highlighting the need for further research into ASR systems capable of handling disordered conversational speech effectively.

Listener effort quantifies clinically meaningful progression of dysarthria in people living with amyotrophic lateral sclerosis.

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative motor neuron disease that causes progressive muscle weakness. Progressive bulbar dysfunction causes dysarthria and thus social isolation, reducing quality of life. The Everything ALS Speech Study obtained longitudinal clinical information and speech recordings from 292 participants. In a subset of 120 participants, we measured speaking rate (SR) and listener effort (LE), a measure of dysarthria severity rated by speech pathologists from recordings. LE intra- and inter-rater reliability was very high (ICC 0.88 to 0.92). LE correlated with other measures of dysarthria at baseline. LE changed over time in participants with ALS (slope 0.77 pts/month; p<0.001) but not controls (slope 0.005 pts/month; p=0.807). The slope of LE progression was similar in all participants with ALS who had bulbar dysfunction at baseline, regardless of ALS site of onset. LE could be a remotely collected clinically meaningful clinical outcome assessment for ALS clinical trials.

Amyotrophic Lateral Sclerosis-Bulbar Dysfunction Index-Remote: Test-Retest and Interrater Reliability of Candidate Items.

PURPOSE: The primary aim of this study was to establish the reliability of candidate items as a step in the development of the Amyotrophic Lateral Sclerosis-Bulbar Dysfunction Index-Remote (ALS-BDI-Remote), a novel tool being developed for the detection and monitoring of bulbar signs and symptoms in remote settings. METHOD: The set of candidate items included 40 items covering three domains: cranial nerve examination, auditory-perceptual evaluation, and functional assessment. Forty-eight participants diagnosed with ALS and exhibiting a range of bulbar disease severity were included. Data collection for each participant took place on Zoom over three sessions. During Session 1, the participants were instructed to adjust their Zoom settings and to optimize their recording environment (e.g., lighting, background noise). Their cognition and eating were screened to determine their ability to follow instructions and their eligibility to perform the swallowing and chewing tasks. During Session 2, two speech-language pathologists (SLPs) administered the tool consecutively to determine the items' interrater reliability. During Session 3, one of the SLPs readministered the tool within 2 weeks of Session 1 to assess test-retest reliability. The reliability of each item was estimated using weighted kappa and the percentage of agreement. To be considered reliable, the items had to reach a threshold of 0.5 weighted kappa or 80% percentage agreement (if skewed distribution of the scores) for both interrater and test-retest reliability. RESULTS: In total, 33 of the 40 candidate items reached the reliability cutoff for both reliability analyses. All assessment domains included reliable items. Items requiring very good visualization of structures or movements were generally less reliable. CONCLUSIONS: This study resulted in the selection of reliable items to be included in the next version of the ALS-BDI-Remote, which will undergo psychometric evaluation (reliability, validity, and responsiveness analyses). Additionally, the results contributed to our understanding of the remote administration of SLP assessments for telehealth applications.