Mapping Artificial Intelligence Research in Aphasia: An AI-Assisted Scoping Review and Semantic Network Analysis
Abstract
Purpose: A shared view of artificial intelligence (AI) research in aphasia is needed to locate related work across clinical and computing publications. This scoping review mapped research topics and connections, AI method families over time, and reported evaluation and participant characteristics relevant to clinical applicability. Method: Seven sources were searched for publications from January 2010 through June 2026. Two large language models (LLMs) independently screened records and extracted information from publication abstracts. Disagreements within and across LLMs were manually reviewed and resolved. To check screening accuracy, three human reviewers independently screened a sample without seeing model responses. Semantic network analysis of the abstracts identified research communities and connecting studies. Results: LLM-assisted screening identified 412 primary studies. Its decisions agreed with the human majority on 91% of 145 sampled records. Network analysis identified 12 communities, with partial separation between clinical and computing publications and general separation between post-stroke aphasia and primary progressive aphasia studies. Deep learning and generative AI first appeared in clinical and computing publications in the same respective years and represented larger proportions of computing publications. Abstracts often lacked key evaluation and participant information: only 28 abstracts (7%) reported independent model testing, 21 (5%) clinical evaluation, and 34 (8%) aphasia severity. Conclusions: The map provides entry points for exploring related research, and the workflow offers a foundation for future evidence syntheses in communication disorders. Clearer reporting of evaluation and participant characteristics could improve communication across disciplines and help readers identify relevant publications.
Related
- Artificial Intelligence-Assisted Systems Biology Reconstruction of Diabetic Bone Marrow Dysfunction
- Contribution of artificial intelligence research to the development of inclusive applications for people with disabilities: A scoping review (2015–2025)
- Hybrid Semantic-Lexical Retrieval for Source-to-CDE Mapping in the NCI caDSR
- Mapping Gene Expression to an Interpretable Semantic Space
- Retrospective harmonization of neuropsychiatric questionnaires in a large multi-site research consortium using expert mappings and semantic textual similarity
BibTeX
@misc{qiu2026mapping,
title = {{Mapping Artificial Intelligence Research in Aphasia: An AI-Assisted Scoping Review and Semantic Network Analysis}},
author = {Mengyang Qiu and Siena Sun and Sara NassrElgrgawi and Zirui Feng},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/vuztw_v1},
url = {https://osf.io/preprints/psyarxiv/vuztw_v1/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Mapping Artificial Intelligence Research in Aphasia: An AI-Assisted Scoping Review and Semantic Network Analysis
AU - Mengyang Qiu
AU - Siena Sun
AU - Sara NassrElgrgawi
AU - Zirui Feng
PY - 2026
DA - 2026/09/16
PB - PsyArXiv
DO - 10.31234/osf.io/vuztw_v1
UR - https://osf.io/preprints/psyarxiv/vuztw_v1/
AB - Purpose: A shared view of artificial intelligence (AI) research in aphasia is needed to locate related work across clinical and computing publications. This scoping review mapped research topics and connections, AI method families over time, and reported evaluation and participant characteristics relevant to clinical applicability. Method: Seven sources were searched for publications from January 2010 through June 2026. Two large language models (LLMs) independently screened records and extracted information from publication abstracts. Disagreements within and across LLMs were manually reviewed and resolved. To check screening accuracy, three human reviewers independently screened a sample without seeing model responses. Semantic network analysis of the abstracts identified research communities and connecting studies. Results: LLM-assisted screening identified 412 primary studies. Its decisions agreed with the human majority on 91% of 145 sampled records. Network analysis identified 12 communities, with partial separation between clinical and computing publications and general separation between post-stroke aphasia and primary progressive aphasia studies. Deep learning and generative AI first appeared in clinical and computing publications in the same respective years and represented larger proportions of computing publications. Abstracts often lacked key evaluation and participant information: only 28 abstracts (7%) reported independent model testing, 21 (5%) clinical evaluation, and 34 (8%) aphasia severity. Conclusions: The map provides entry points for exploring related research, and the workflow offers a foundation for future evidence syntheses in communication disorders. Clearer reporting of evaluation and participant characteristics could improve communication across disciplines and help readers identify relevant publications.
ER -