O impacto da escolaridade e da doença de alzheimer na produção oral de narrativa no envelhecimento : estudo da micro e macroestrutura com análise automática de textos

Abstract

Narrative is a frequent type of text (JERÔNIMO; HÜBNER, 2014). Producing a text is a strategic task of cognitive planning, which is based on the construction of a cognitive model of the text in episodic memory (EM) (VAN DIJK, 2010). In Alzheimer’s disease (AD), discoursive linguistic abilities are impaired since the very beginning of the continuum of the dementia (SOARES; BRANDÃO; LACERDA, 2012). In this sense, automated speech analysis has the potential to distinguish between healthy and cognitively impaired groups, helping in the establishment of early diagnosis (TOLEDO, et. al., 2017). The level of education can also interfere in the quality of text production (JERÔNIMO, 2016; STERN, 2012). Hence, this master’s thesis investigates whether there are differences in oral production of micro and macrostructure of narratives in the comparison between groups of healthy and AD diagnosed elders. It also correlates working (WM) and episodic memory (EM) and linguistic data for the groups. To complement linguistic data, automatic text analysis is conducted in this study. To do so, this thesis is divided in two studies: Study 1 compares the production of micro and macrostructural levels of a narrative made by two groups of healthy elders (divided into high and low educational levels), correlating linguistic data with scores from WM and EM tasks. The performances of 28 high educational level (≥9 years of education) healthy elders (mean age = 69,46) (HHE), and 35 low educational level (between 2 and 8 years of education) healthy elders (mean age = 72,20) (HLE) are analyzed; Study 2 investigates the same comparisons reported in Study 1, but focusing on dementia. A low educational level control group (the same one described in Study 1) (HLE, now CG) is compared to a group of 12 low educational level AD-diagnosed elders (mean age = 74,58) (DA). The linguistic task proposed is an oral production of a narrative based on a sequence of pictures, The Dog Story (LE BOEUF, 1976). The neuropsychological tasks that measure EM and WM are, respectively, Verbal Learning Task, from BALE (Bateria de Avaliação da Linguagem no Envelhecimento) (HÜBNER, et. al., accepted for publication, adapted from BUSCHKE, GROBER, 2012) and Digit Span (subtest from NEUPSILIN, by FONSECA et al., 2009). To the analysis of the microstructure, 23 linguistic features are extracted from the software Coh-Metrix Dementia (ALUÍSIO; CUNHA; SCARTON, 2016). To the analysis of macrostructure the protocol from BALE is adopted. Both quantitative and qualitative analyses are conducted. Main results indicate influence of educational level on performance in the linguistic task, especially in the construction of the macrostructure, which means that low educational level elders had problems with the construction of the different steps of the narrative, included extra information, among other aspects. There were no significant differences between healthy participants in the neuropsychological tasks, even though positive correlations were identified between them and the construction of micro and macrostructure. Differences were also identified in the comparison between CG and AD in the EM task, which may indicate impairment in EM due to dementia. The main linguistic differences between CG and AD are related to the production of fragments of a sentence, higher incidence of content words rather than function words, higher incidence of pauses and hesitations in the narratives produced by ADs. It is expected that the data presented here contribute to foster the use of discourse in AD assessment, as well as to enhance automated speech analysis tools, making it possible to early diagnose dementias and improve life quality to the elderly.