The DOI matches its Crossref record; 138 of the 138 references it cites were found in a registry or an index; 13 of the 83 sources its claims cite were checked for support.
Early Identification of Language Disorders Using Natural Language Processing and Machine Learning: Challenges and Emerging Approaches
Open accessAbstract
Purpose: Recent advances in artificial intelligence provide opportunities to capture and represent complex features of human language in a more automated manner, offering potential means of improving the efficiency of language assessment. This review article presents computerized approaches for the analysis of narrative language and identification of language disorders in children. Method: We first describe the current barriers to clinicians' use of language sample analysis, narrative language sampling approaches, and the data processing stages that precede analysis. We then present recent studies demonstrating the automated extraction of linguistic features and identification of developmental language disorder using natural language processing and machine learning. We explain how these tools operate and emphasize how the decisions made in construction impact their performance in important ways, especially in the analysis of child language samples. We conclude with a discussion of major challenges in the field with respect to bias, access, and generalizability across settings and applications. Conclusion: Given the progress that has occurred over the last decade, computer-automated approaches offer a promising opportunity to improve the efficiency and accessibility of language sample analysis and expedite the diagnosis and treatment of language disorders in children.
Overview
CoChat Summary
TL;DR
This is a review article on how natural language processing (NLP) and machine learning (ML) can automate parts of language sample analysis (LSA) to help identify developmental language disorder (DLD) in children. The paper argues these methods are promising, but their use is constrained by child-language performance, generalizability, interpretability, bias, and the need for human oversight.
Key findings
- LSA is described as a robust, ecologically valid assessment approach, but the paper says its data collection, transcription, annotation, and analysis are time consuming and unreliable; clinicians also report time constraints, lack of training, and lack of computing resources as barriers.
- The review reports that automated approaches can reduce scoring time: one IPSyn system reduced scoring from 30 minutes per 100 utterances to less than 5 minutes, and improved automated IPSyn accuracy to above 95%.
- Across reviewed classifiers, DLD prediction accuracy ranged from 40% to 90% depending on task and sample; the paper says BERT and Bayesian classifiers are currently best for identifying language disorder from children’s narrative language samples.
- Specific reported performance examples include a Bayesian network classifier with F1 = 0.914, a logistic ridge regression classifier with 81% accuracy for severe DLD versus typical development but 68% for less severe DLD, decision trees with F1 = 79% for DLD vs typical development, and F1 = 52% for DLD vs autism.
- For narrative macrostructure scoring, the paper reports near perfect agreement (QWK > 0.9) for all elements except consequence (QWK = 0.79) using BERT, and for narrative microstructure an automated approach had QWK values ranging from 0.74 to 0.89 and outperformed non-expert scorers on 4 of 6 measures.
Methods
- This is a review article synthesizing prior work on automated language sample analysis for children’s narrative language and DLD identification; no new dataset was collected or analyzed.
- The paper reviews automated approaches at multiple LSA stages, including transcription/annotation, feature extraction (microstructure and macrostructure), and ML classification.
- It summarizes reported models and evaluation procedures from the literature, including leave-one-out, hold-out, and k-fold cross-validation, and compares algorithms such as logistic regression, Bayesian networks, Naïve Bayes, SVM, decision trees, random forests, neural networks, and BERT.
Limitations
- The paper explicitly states that automated LSA models still require human oversight and that users cannot completely rely on fully automated end-to-end pipelines.
- It states that many NLP tools are not optimized for child language and often require significant manual correction; ASR systems trained on adult speech can have higher-than-acceptable error rates for children’s speech.
- The article is limited to narrative language analysis in children; this is stated in the conclusion.
- Evident from the text: the review draws on studies with small samples and limited demographic reporting, so its claims about performance and fairness depend on a literature base that the paper itself says is not yet sufficiently representative.
Model-written summary of a shortened version of the paper’s full text (it is longer than we read). Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
9 Valid · 5 Not valid · 47 Not checked · 64 Own results and interpretations
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“However, despite their accessibility, objectivity, and cost effectiveness, few SLPs (2%-7%) reported using computer-based approaches (McLeod & Baker, 2014; Pavelko et al., 2016).”
What the sources say
- Could not verifycites McLeod & Baker, 2014 sourceExperimental — accuracy not yet measured
The supplied passage is bibliographic information, not a report of SLPs’ practices. Searches of our indexed copy for computer-based approaches, the 2%–7% figure, and related assessment terms found no passage addressing the claim; it therefore cannot be verified from our copy. (LLM, conf=0.70)
- Not checked — no full text of the sourcecites Pavelko et al., 2016 source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“DLD often co-occurs with other communication and developmental disorders such as autism spectrum disorder (ASD) and attention deficit hyperactivity disorder (ADHD; Bishop et al., 2017).”
What the sources say
- Partly supportedcites Bishop et al., 2017 sourceExperimental — accuracy not yet measured
Statement 9 (Results and discussion, Consensus statements; page unavailable) says ADHD and speech problems can co-occur with DLD. But Statement 6 places ASD among differentiating conditions, and Statement 7 reserves the term DLD for language disorder without a known differentiating condition. The paper therefore does not support naming ASD as an example of a disorder co-occurring with DLD under its terminology. (LLM, conf=0.91)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Moreover, limited access to practitioners who can provide clinical diagnoses based on LSA outcomes (e.g., Speech-Language Pathologists [SLPs]) due to costs and poor availability in areas where service does not meet community demand (Lim et al., 2017; McAllister et al., 2011) can further drive inequity and discrimination (O'Callaghan et al., 2005; Verdon et al., 2011).”
What the sources say
- Not checked — no full text of the sourcecites Lim et al., 2017 source
- Could not verifycites McAllister et al., 2011 sourceExperimental — accuracy not yet measured
The available indexed passages are bibliographic metadata (Abstract listing, no page), including the title and a link to an accepted manuscript, rather than the paper’s discussion. Searches for costs, practitioner availability, waiting lists, and LSA-based diagnoses found no passage addressing the attributed claim. The decisive text may be absent from the indexed copy. (LLM, conf=0.72)
- Not checked — no full text of the sourcecites O'Callaghan et al., 2005 source
plainstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Algorithmic bias can arise at nearly any stage of the clinical ML workflow, resulting in inequitable diagnosis and treatment outcomes across sub-populations (Chen et al., 2023).”
What the sources say
- Partly supportedcites Chen et al., 2023 sourceExperimental — accuracy not yet measured
The Discussion (page unavailable) identifies sampling bias, methodological flaws, and limited technical availability as problems for psychiatric diagnostic classifiers, and warns that underrepresentation may exacerbate healthcare inequalities. But its limitations paragraph explicitly says the study focused on diagnosis, not treatment-outcome prediction. It also does not establish that bias arises at nearly every stage of the clinical ML workflow. (LLM, conf=0.88)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Current accuracy levels are such that users cannot completely rely on fully automated end-to-end pipelines (i.e., automation of everything from sample transcription to diagnosis) without human oversight.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
40 more claims
Create a free account to see all 125 claims and every source verdict.
- It may state two separate claims as one. 50
- Probably not a substantial claim. 53
- The quote only partly states this claim. 1
Citations
Cites
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173
Glove: Global Vectors for Word Representation
Pennington, J., Socher, R., & Manning, C. (2014) · Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.3115/v1/d14-1162 · Cited by 34,071
(2003) · Journal of Machine Learning Research · doi:10.5555/944919.944937 · Cited by 32,337
The weirdest people in the world?
Henrich, J., Heine, S. J., & Norenzayan, A. (2010) · Behavioral and Brain Sciences · doi:10.1017/s0140525x0999152x · Cited by 12,641
Blei, D. M., Ng, A. Y., & Jordan, M. I. (2002) · Advances in Neural Information Processing Systems 14 · doi:10.7551/mitpress/1120.003.0082 · Cited by 12,223
An Introduction to Statistical Learning
James, G., Witten, D., Hastie, T., & Tibshirani, R. (2013) · Springer Texts in Statistics · doi:10.1007/978-1-4614-7138-7 · Cited by 12,106
A Study of Cross-Validation and Bootstrap for Accuracy Estimation and Model Selection
Kohavi, R. (1995) · Zenodo · doi:10.5281/zenodo.19712698 · Cited by 10,909
Pattern Recognition and Machine Learning
(2006) · doi:10.1007/978-0-387-45528-0 · Cited by 9,824
Dissecting racial bias in an algorithm used to manage the health of populations
Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019) · Science · doi:10.1126/science.aax2342 · Cited by 7,137
Research works printed but not linked: 3
- Gini, C. (1936). On the measure of concentration with special reference to income and statistics. Colorado College Publication, 208, 73–79.
- Hansson, K., Sandgren, O., & Sahlén, B. (2014). Changing labels for a concept in change. Comment to Bishop, D. Ten questions about terminology for children with unexplained language problems. International Journal of Language & Communication Disorders, 49, 407–408.
- Zeigler, K., & Camarota, A. (2019). 67.3 million in the United States spoke a foreign language at home in 2018. Center for Immigration Studies. https://cis.org/Report/673-Million-United-States-Spoke-Foreign-Language-Home-2018#5
Cited by
Winter, P., Bornman, J., & van der Linde, J. (2026) · International Journal of Speech-Language Pathology · doi:10.1080/17549507.2026.2630839 · Cited by 2
Georgiou, G. P. (2025) · Annals of Biomedical Engineering · doi:10.1007/s10439-025-03820-0 · Cited by 3
“The detection has been tested in a variety of models, primarily supervised ML [8], including Bayesian networks, naive Bayes, support vector machines (SVM), neural networks, and decision tree models [9], alongside unsupervised ML [10].”
Developmental Sentence Scoring for Preschool Language Sample Analysis: A Psychometric Update
Ratner, N. B., MacWhinney, B., Hong, Y., & Yang, J. S. (2026) · American Journal of Speech-Language Pathology · doi:10.1044/2025_ajslp-24-00525 · Cited by 2
A Deep Residual Recurrent Learning Approach for the Identification of Learning Disability Disorders
Devi, T. S., Gopalakrishnan, S., & Chand, S. R. (2026) · International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT) · doi:10.1109/idciot67589.2026.11455681 · Cited by 1
Metadata
- Authors
- Lammert, J. M., Roberts, A. C., McRae, K., Batterink, L. J., & Butler, B. E.
- Affiliations
- Western University (CA)
- Venue
- Journal of Speech, Language, and Hearing Research
- Published
- 2025-02-04
- Volume
- 68
- Issue
- 2
- Pages
- 705-718
- ISSN
- 1092-4388 (ISSN-L)
- 1558-9102
- OpenAlex
- W4406206936
- Semantic Scholar
- 61769c5bb5d5ac668c69d146385ccf09fd07ced4
- Cited by
- 17
- References
- 109
- Citations by year
- 2026: 14 · 2025: 3
- FWCI
- 4.20
- Type
- journal article
- Language
- en
- Open access
- green
- dx.doi.org
- Licence
- other-oa
- Locations
- Journal of Speech Language and Hearing Research (publishedVersion)
- PubMed (publishedVersion)
- PsyArXiv (OSF Preprints) (submittedVersion, cc-by)
- Topics
- Social Sciences › Psychology › Developmental and Educational Psychology › Language Development and Disorders
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Health Sciences › Health Professions › Speech and Hearing › Voice and Speech Disorders
- Keywords
- specific language impairment, early detection, language sample analysis, narrative language, children, generalizability, natural language processing, machine learning
- MeSH terms
- Machine Learning, Child, Humans, Language Development Disorders, Language Disorders, Language Tests, Natural Language Processing, Early Diagnosis
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-01, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-01 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Early Identification of Language Disorders Using Natural Language Processing and Machine Learning: Challenges and Emerging Approaches. https://app.cochat.ai/papers/10.1044/2024_jslhr-24-00515
- The paper
- Lammert, J. M., Roberts, A. C., McRae, K., Batterink, L. J., & Butler, B. E. (2025). Early Identification of Language Disorders Using Natural Language Processing and Machine Learning: Challenges and Emerging Approaches. Journal of Speech, Language, and Hearing Research. https://doi.org/10.1044/2024_jslhr-24-00515
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.