Aphasia is a common speech and language disorder, typically caused by a brain injury or a stroke, that affects millions of people worldwide. Detecting and assessing Aphasia in patients is a difficult, time-consuming process, and numerous attempts to automate it have been made, the most successful using machine learning models trained on aphasic speech data. Like in many medical applications, aphasic speech data is scarce and the problem is exacerbated in so-called "low resource" languages, which are, for this task, most languages excluding English. We attempt to leverage available data in English and achieve zero-shot aphasia detection in low-resource languages such as Greek and French, by using language-agnostic linguistic features. Current cross-lingual aphasia detection approaches rely on manually extracted transcripts. We propose an end-toend pipeline using pre-trained Automatic Speech Recognition (ASR) models that share cross-lingual speech representations and are fine-tuned for our desired low-resource languages. To further boost our ASR model's performance, we also combine it with a language model. We show that our ASR-based end-toend pipeline offers comparable results to previous setups using human-annotated transcripts.
In this paper we present a web-based data collection method designed to elicit narrative discourse from adults with and without language impairments, both in an in-person set up and remotely. We describe the design, methodological considerations and technical requirements regarding the application development, the elicitation tasks, materials and guidelines, as well as the implementation of the assessment procedure. To investigate the efficacy of remote elicitation of narrative discourse with the use of the technology-enhanced method presented here, a pilot study was conducted, aiming to compare narratives elicited remotely to narratives collected in an in-person elicitation mode from ten unimpaired adults, using a within-participants research design. In the remote elicitation setting, each participant performed the tasks of a narrative elicitation protocol via the web application in their own environment, with the assistance of an investigator in the context of a virtual meeting (video conferencing). In the in-person elicitation setting, the participant was in the same environment with the investigator, who administered the tasks using the web application. Data were manually transcribed, and transcripts were processed with Natural Language Processing (NLP) tools. Linguistic features representing key measures of spoken narrative discourse were automatically calculated: linguistic productivity, content richness, fluency, syntactic complexity at clausal and inter-clausal level, lexical diversity, and verbal output. The results show that spoken narratives produced by the same individuals in the two different experimental settings do not present significant differences regarding the linguistic variables analyzed, in sixty six out of seventy statistical tests. These results indicate that the presented web-based application is a feasible method for the remote collection of spoken narrative discourse from adults without language impairments in the context of online assessment.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.