Aphasia is an acquired language disorder, often resulting from a stroke, affecting nearly 580,000 people Europe alone each year (Huber et al., 2013). Depending on the type and severity, people with aphasia suffer, in varying degrees, from the impairment of one or several of the four communication modalities. To choose an appropriate therapy for a patient, the extent of the aphasia at hand has to be diagnosed. In Germany and other countries this is done using the Aachen Aphasia Test (AAT). The AAT consists of a series of tests, requiring the patient to talk, read and write over the course of up to two hours. The AAT results then have to be evaluated by a speech and language therapist, which takes around 6 hours. In order to further objectify the manual diagnosis and speed up the process, a digital support system would be highly valuable for the clinical field. To facilitate such a system, we have collected, cleaned and processed real-life clinical aphasia data, coming from AAT diagnosis sessions. Each dataset consists of speech data, a transcript and rich linguistic AAT annotations. In this paper, we report on both challenges and early results in working with the (raw) clinical aphasia data.
@InProceedings{KOHLSCHEIN18.2, author = {Christian Kohlschein and Daniel Klischies}, title = {Automatic Processing of Clinical Aphasia Data collected during Diagnosis Sessions: Challenges and Prospects}, booktitle = {Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)}, year = {2018}, month = {may}, date = {7-12}, location = {Miyazaki, Japan}, editor = {Dimitrios Kokkinakis}, publisher = {European Language Resources Association (ELRA)}, address = {Paris, France}, isbn = {979-10-95546-26-9}, language = {english} }
Summary of the paper