Towards a Method for Synthetic Generation of Persons with Aphasia Transcripts Jason M. Pittman1, Anton Phillips Jr.2, Yesenia Medina-Santos2, Brielle C. Stark2 1University of Maryland Global Campus 2Indiana University Bloomington, Department of Speech, Language and Hearing Sciences ABSTRACT In aphasia research, Speech-Language Pathologists (SLPs) devote extensive time to manually coding speech samples using Correct Information Units (CIUs), a measure of how informative an individual sample of speech is. Developing automated systems to recognize aphasic language is limited by data scarcity. For example, only about 600 transcripts are available in AphasiaBank yet billions of tokens are used to train large language models (LLMs). In the broader field of machine learning (ML), researchers increasingly turn to synthetic data when such are sparse. Therefore, this study constructs and validates two methods to generate synthetic transcripts of the AphasiaBank Cat Rescue picture description task. One method leverages a procedural programming approach while the second uses Mistral 7b Instruct and Llama 3.1 8b Instruct LLMs. The methods generate transcripts across four severity levels (Mild, Moderate, Severe, Very Severe) through word dropping, filler insertion, and paraphasia substitution. Overall, we found, compared to human-elicited transcripts, Mistral 7b Instruct best captures key aspects of linguistic degradation observed in aphasia, showing realistic directional changes in NDW, word count, and word length amongst the synthetic generation methods. Based on the results, future work should plan to create a larger dataset, fine-tune models for better aphasic representation, and have SLPs assess the realism and usefulness of the synthetic transcripts. Keywords: aphasia, synthetic data, natural language processing, machine learning Introduction Per Nicholas and Brookshire (1993), coding Correct Information Units (CIUs) involves transcribing a connected speech sample verbatim, counting all intelligible words, and then identifying each word that is intelligible, accurate, relevant, and informative about the topic as a CIU--excluding fillers, repetitions, and tangential remarks. From these counts, clinicians calculate the percentage of CIUs and CIUs per minute to quantify communicative informativeness and efficiency.

Synthetic transcript generation is critical in contact center domains, where privacy and data scarcity limit model training and evaluation. Unlike prior synthetic dialogue generation work on open-domain or medical dialogues, contact center conversations are goal-oriented, role-asymmetric, and behaviorally complex, featuring disfluencies, ASR noise, and compliance-driven agent actions. In deployments where transcripts are unavailable, standard pipelines still yield derived call attributes such as Intent Summaries, Topic Flow, and QA Evaluation Forms. We leverage these as supervision signals to guide generation. To assess the quality of such outputs, we introduce a diagnostic framework of 18 linguistically and behaviorally grounded metrics for comparing real and synthetic transcripts. We benchmark four language-agnostic generation strategies, from simple prompting to characteristic-aware multi-stage approaches, alongside reference-free baselines. Results reveal persistent challenges: no method excels across all traits, with notable deficits in disfluency, sentiment, and behavioral realism. Our diagnostic tool exposes these gaps, enabling fine-grained evaluation and stress testing of synthetic dialogue across languages.

Accurate and comprehensive clinical documentation is crucial for delivering high-quality healthcare, facilitating effective communication among providers, and ensuring compliance with regulatory requirements. However, manual transcription and data entry processes can be time-consuming, error-prone, and susceptible to inconsistencies, leading to incomplete or inaccurate medical records. This paper proposes a novel approach to augment clinical documentation by leveraging synthetic data generation techniques to generate realistic and diverse clinical transcripts. We present a methodology that combines state-of-the-art generative models, such as Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), with real-world clinical transcript and other forms of clinical data to generate synthetic transcripts. These synthetic transcripts can then be used to supplement existing documentation workflows, providing additional training data for natural language processing models and enabling more accurate and efficient transcription processes. Through extensive experiments on a large dataset of anonymized clinical transcripts, we demonstrate the effectiveness of our approach in generating high-quality synthetic transcripts that closely resemble real-world data. Quantitative evaluation metrics, including perplexity scores and BLEU scores, as well as qualitative assessments by domain experts, validate the fidelity and utility of the generated synthetic transcripts. Our findings highlight synthetic data generation's potential to address clinical documentation challenges, improving patient care, reducing administrative burdens, and enhancing healthcare system efficiency.

End-to-end automatic speech translation (AST) relies on data that combines audio inputs with text translation outputs. Previous work used existing large parallel corpora of transcriptions and translations in a knowledge distillation (KD) setup to distill a neural machine translation (NMT) into an AST student model. While KD allows using larger pretrained models, the reliance of previous KD approaches on manual audio transcripts in the data pipeline restricts the applicability of this framework to AST. We present an imitation learning approach where a teacher NMT system corrects the errors of an AST student without relying on manual transcripts. We show that the NMT teacher can recover from errors in automatic transcriptions and is able to correct erroneous translations of the AST student, leading to improvements of about 4 BLEU points over the standard AST end-to-end baseline on the English-German CoVoST-2 and MuST-C datasets, respectively. Code and data are publicly available.\footnote{\url{https://github.com/HubReb/imitkd_ast/releases/tag/v1.1}}