?
Кластеризация лексики для решения задачи диагностики афазии
In this article an algorithm of automatic vocabulary clustering is implemen- ted to solve the problem of creating a computer model for aphasia diagnostics based on the results of the verbal fluency test. This test involves generating the maximum number of words that fit a specified category within a given time limit. When analyzing test results, researchers often examine both quantitative and qualitative metrics: the total number of correct responses and the number of “switches” – transitions from one subcategory of lexicon to another. The subcategories are presented as groups of lexemes that are organized based on some common attributes. They are usually referred to as clusters and their sizes serve as additional qualitative indicators of test performance. The study hypothesizes that the respondents who were affected by aphasia produced a lower number of responses with fewer switches, but showed similar cluster size, as well as less distinct semantic similarity in and between clusters. The authors employed a rule-based algorithm of automatic vocabulary clustering, proposed in the work of N. Lundin et al. (Psychiatry Research. 2022. Т. 309), along with the materials collected and described in the work of O. Buivolova et al. (Russian Journal of Cognitive Science. 2020. Vol. 7. No. 3). As a result, a functional model for determining aphasia among Russian speakers was created. The findings also indicate that respondents with aphasia tend to give less standardized answers, forming fewer subcategories, but with a greater associative score within clusters. Additionally, comparative analysis with data from The Russian National Corpus revealed that respondents of both groups tend to provide associations, that are typical and standardized for native Russian speakers.