Evaluating atypical language in autism using automated language measures

Salem, Alexandra C.; MacFarlane, Heather; Adams, Joel Robert; Lawley, Grace O; Dolata, Jill; Bedrick, Steven; Fombonne, Éric

doi:10.1038/s41598-021-90304-5

Cited by 19 publications

(31 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Nevertheless, several research implications of our findings and future directions are worth noting. First, using natural language processing approaches, our group has generated several automatic discourse measures that, when applied to transcripts of ADOS-2 in the same sample have shown an ability to differentiate youth with ASD from controls, both in isolation and when taken in combination ( Salem et al, 2021 ). A logical next step will be for us to compare the levels of accuracy that can be achieved, on the same subjects, by voice analysis or language analysis only, and to evaluate if combining voice and language analysis would result in gains of accuracy of predicting diagnostic status.…”

Section: Discussionmentioning

confidence: 99%

Quantifying Voice Characteristics for Detecting Autism

2021

Self Cite

View full text Add to dashboard Cite

The presence of prosodic anomalies in autistic is recognized by experienced clinicians but their quantitative analysis is a cumbersome task beyond the scope of typical pen and pencil assessment. This paper proposes an automatic approach allowing to tease apart various aspects of prosodic abnormalities and to translate them into fine-grained, automated, and quantifiable measurements. Using a harmonic model (HM) of voiced signal, we isolated the harmonic content of speech and computed a set of quantities related to harmonic content. Employing these measures, along with standard speech measures such as loudness, we successfully trained machine learning models for distinguishing individuals with autism from those with typical development (TD). We evaluated our models empirically on a task of detecting autism on a sample of 118 youth (90 diagnosed with autism and 28 controls; mean age: 10.9 years) and demonstrated that these models perform significantly better than a chance model. Voice and speech analyses could be incorporated as novel outcome measures for treatment research and used for early detection of autism in preverbal infants or toddlers at risk of autism.

show abstract

Section: Discussionmentioning

confidence: 99%

Quantifying Voice Characteristics for Detecting Autism

2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…In the present study, we aimed to confirm the performance of the automated voice and language measurements as previously established by Asgari et al (2021) and Salem et al (2021), and further, to evaluate any potential gains derived from their combination. Specifically, our approach is to estimate the accuracy of a combined voice and language model in correctly classifying ASD status in a mixed sample of already-diagnosed children and a non-autistic comparison group, approaching this work with a streamlined, purposeful inclusion of established voice and language measures.…”

Section: Introductionmentioning

confidence: 88%

“…For example, Alhanai et al (2017) found that combining audio and text features increased detection of cognitive impairment in a large, longitudinal population study over both baseline demographic performance and over feature grouping. When applied to autism, NLP methods have also been used efficiently to characterize language profiles (Chojnicka & Wawer, 2020; Salem et al, 2021). Likewise, voice analyses have been performed showing discriminant prosodic patterns in autistic versus non‐autistic participant groups (Asgari et al, 2021; Kiss et al, 2012; Li et al, 2019).…”

Section: Introductionmentioning

confidence: 99%

“…In previous work, we have used NLP methodologies and established their ability to differentiate between diagnostic groups (Adams et al, 2021; MacFarlane et al, 2017; Salem et al, 2021; van Santen et al, 2013). Salem et al (2021) examined seven automated language measures (ALMs) and found them able to accurately classify already‐diagnosed children into ASD and non‐ASD groups even after controlling for age and IQ.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Combining voice and language features improves automated autism detection

et al. 2022

Self Cite

View full text Add to dashboard Cite

Variability in expressive and receptive language, difficulty with pragmatic language, and prosodic difficulties are all features of autism spectrum disorder (ASD). Quantifying language and voice characteristics is an important step for measuring outcomes for autistic people, yet clinical measurement is cumbersome and costly. Using natural language processing (NLP) methods and a harmonic model of speech, we analyzed language transcripts and audio recordings to automatically classify individuals as ASD or non‐ASD. One‐hundred fifty‐eight participants (88 ASD, 70 non‐ASD) ages 7 to 17 were evaluated with the autism diagnostic observation schedule (ADOS‐2), module 3. The ADOS‐2 was transcribed following modified SALT guidelines. Seven automated language measures (ALMs) and 10 automated voice measures (AVMs) for each participant were generated from the transcripts and audio of one ADOS‐2 task. The measures were analyzed using support vector machine (SVM; a binary classifier) and receiver operating characteristic (ROC). The AVM model resulted in an ROC area under the curve (AUC) of 0.7800, the ALM model an AUC of 0.8748, and the combined model a significantly improved AUC of 0.9205. The ALM model better detected ASD participants who were younger and had lower language skills and shorter activity time. ASD participants detected by the AVM model had better language profiles than those detected by the language model. In combination, automated measurement of language and voice characteristics successfully differentiated children with and without autism. This methodology could help design robust outcome measures for future research. Lay Summary People with autism often struggle with communication differences which traditional clinical measures and language tests cannot fully capture. Using language transcripts and audio recordings from 158 children ages 7 to 17, we showed that automated, objective language and voice measurements successfully predict the child's diagnosis. This methodology could help design improved outcome measures for research.

show abstract

“…To date, a handful of studies have used computational or word frequency-based approaches to examine language produced during the ADOS-2 in autistic school-age youth (in addition to research using qualitative coding [ 81 ]). These studies focused primarily on lexico-semantic aspects of language including disfluencies [ 33 ], sentiment and linguistic abstraction [ 82 ], nouns versus cognitive process words [ 30 ], latent semantic similarity [ 83 ], number of word roots and content maze repetition [ 84 ], and acoustic-prosodic features [ 85 – 87 ]. However, children’s use of social words more broadly during the interview sections of the ADOS-2 has not been explored, and critically, only two prior studies included large enough samples of autistic girls or women to examine potential sex differences [ 30 , 33 ].…”

Section: Introductionmentioning

confidence: 99%

Friend matters: sex differences in social language during autism diagnostic interviews

et al. 2022

View full text Add to dashboard Cite

Background Autistic individuals frequently experience social communication challenges. Girls are diagnosed with autism less often than boys even when their symptoms are equally severe, which may be due to insufficient understanding of the way autism manifests in girls. Differences in the behavioral presentation of autism, including how people talk about social topics, could contribute to these persistent problems with identification. Despite a growing body of research suggesting that autistic girls and boys present distinct symptom profiles in a variety of domains, including social attention, friendships, social motivation, and language, differences in the way that autistic boys and girls communicate verbally are not yet well understood. Closely analyzing boys’ and girls’ socially-focused language during semi-structured clinical assessments could shed light on potential sex differences in the behavioral presentation of autistic individuals that may prove useful for identifying and effectively supporting autistic girls. Here, we compare social word use in verbally fluent autistic girls and boys during the interview sections of the ADOS-2 Module 3 and measure associations with clinical phenotype. Methods School-aged girls and boys with autism (N = 101, 25 females; aged 6–15) were matched on age, IQ, and parent/clinician ratings of autism symptom severity. Our primary analysis compared the number of social words produced by autistic boys and girls (normalized to account for differences in total word production). Social words are words that make reference to other people, including friends and family. Results There was a significant main effect of sex on social word production, such that autistic girls used more social words than autistic boys. To identify the specific types of words driving this effect, additional subcategories of friend and family words were analyzed. There was a significant effect of sex on friend words, with girls using significantly more friend words than boys. However, there was no significant main effect of sex on family words, suggesting that sex differences in social word production may be driven by girls talking more about friends compared to boys, not family. To assess relationships between word use and clinical phenotype, we modeled ADOS-2 Social Affect (SA) scores as a function of social word production. In the overall sample, social word use correlated significantly with ADOS-2 SA scores, indicating that participants who used more social words were rated as less socially impaired by clinicians. However, when examined in each sex separately, this result only held for boys. Limitations This study cannot speak to the ways in which social word use may differ for younger children, adults, or individuals who are not verbally fluent; in addition, there were more autistic boys than girls in our sample, making it difficult to detect small effects. Conclusions Autistic girls used significantly more social words than boys during a diagnostic assessment—despite being matched on age, IQ, and both parent- and clinician-rated autism symptom severity. Sex differences in linguistic markers of social phenotype in autism are especially important in light of the late or missed diagnoses that disproportionately affect autistic girls. Specifically, heightened talk about social topics could complicate autism referral and diagnosis when non-clinician observers expect a male-typical pattern of reduced social focus, which autistic girls may not always exhibit.

show abstract

Evaluating atypical language in autism using automated language measures

Cited by 19 publications

References 38 publications

Quantifying Voice Characteristics for Detecting Autism

Quantifying Voice Characteristics for Detecting Autism

Combining voice and language features improves automated autism detection

Friend matters: sex differences in social language during autism diagnostic interviews

Contact Info

Product

Resources

About