Marek Dobeš (CSPS SAS) and Pavol Drotár (TUKE) research the question whether it is possible for a computer algorithm to detect from a sample of writing whether a child may be dysgraphic. The research is supported by the APVV grant “Computer-Aided Decision Support System for Hepatic Encephalopathy” and by the VEGA grant “Computational model of integration of chemosensory and motor modules of C. Elegans neural network”.
Dysgraphia is a writing disorder manifesting in unruly writing or switching letters or other distortions in writing. As writing is a strong part of our culture and in schools much time is devoted to writing, a child with dysgraphia may have problems in academic life. Furthermore, these problems can transform into problems with self-concept.
That is why early diagnostics of dysgraphia is important. Currently, trained professionals carry out dysgraphia diagnostics. That means however that their numbers are limited and it is not possible to test every child. It may happen that a child with dysgraphia may not be diagnosed and may encounter problems at school. Dysgraphia can be wrongly interpreted as laziness or lack of motivation to train writing.
Advances in computer science enable computers to process real-world data and classify them as precisely (and sometimes more) as humans do. Computer algorithms are for example used in diagnostics of tumors from medical scans. Computer is presented with a load of data and it learns to extract important features that enable it to differentiate respective categories.
When screening for dysgraphia a child writes a few sentences on a paper placed on a specialised tablet. The tablet records not only the writing but also how long a child writes, how many times s/he pauses, what is the pen pressure, how is the pen lift and tens of other kinds of data. It is an advantage over a human tester who usualy tests only for a couple of features.
Next the algorithm compares the values of these parameters with values that have been presented in the training. After few seconds it describes the probability with which a child may be dysgraphic. If this probability is sufficiently high, a recommendation may be sent to the parents of a child to see a specialist. In this way it is possible to test much more children and give them the chance to treat this disorder early.
Currently, alternative algorithms are being tested with the accuracy about 80.