O papel do contexto semântico no reconhecimento de marcadores prosódicos da escrita do português brasileiro por leitores de diferentes níveis de escolaridade

Abstract

The present study is aimed to investigate the relationship between the level of reading fluency and the level of education with the recognition of lexical prosodic markers (MPLs) and graphical prosodic markers (MPGs) in the written Brazilian Portuguese language. In addition, it intends to assess the role of context in the recognition of MPGs and MPLs, in order to enable or disable the contextual module of the prosodic marker recognition architecture proposed by Pacheco and Oliveira (2014). Therefore, we observed and analysed subjects reading aloud in different levels of education and their relationship with the understanding of the material read. Fifteen elementary school students from the 6th year participated in the research, 15 high school students from the 3rd year and 15 under graduated students from the 7th semester. Each participant read a text silently at first and then answered a comprehension questionnaire on the laptop. Then, each participant read the text aloud. From reading aloud, we observed the following: speaking time and rate of articulation. From the comprehension questionnaire, we verified the following: response time, percentage of error and percentage of correct answers. To assess the processing of MPLs and MPGs, target phrases under the scope of the MPLs ("screamed", "whispered", "shouted", "said softly", "said quickly" and "said slowly" and MPGs (exclamation point, full stop and question mark) were prepared to occur separately. To investigate the role of context in the processing of MPLs and MPGs, four semantic contexts were chosen: angry, affectionate, joyful and sad which were called emotional contexts, considering that MPLs are associated with attitudes and emotions of speakers, and target phrases under the scope of MPLs and MPGs were placed in texts that covered these four contexts. The texts were recorded in an acoustically treated booth for accurate acoustic analysis through which the following variables: fundamental frequency (F0), intensity and duration were checked. As a result, we observed a directly proportional relationship between the level of education and the rate of articulation of the text and the percentage of correct answers, and an inversely proportional relationship between the level of education, the time from speech and the percentage of error. Regarding the total time spent to answer the questions correctly, there was a statistically significant difference separating the higher education group, with less time spent to answer the questions correctly, from both the elementary school group and the high school group. The MPGs were recognized by the three investigated groups, with statistically significant differences between the three MPGs. However, when these MPGs occurred in the contexts, only the groups with the highest level of fluency and education showed statistically significant differences between the contexts of occurrence of each MPG. As for the MPLs, for the high school and higher education group, we found statistically significant differences in all variables, which separated the different MPLs. These groups also adjusted prosodic parameters to differentiate the contexts of occurrence of each MPL. The elementary school group showed a statistically significant difference only for the intensity variable and did not show statistically significant differences between the contexts of occurrence of the MPLs. The results showed that the more educated readers showed better performance in oral reading (shorter speaking time and higher articulation rate) and in the comprehension test, with less time spent to answer questions correctly and lower percentage of error. The more educated readers were more efficient in differentiating from the MPLs' phonetic point of view. The processing of Brazilian Portuguese writing MPLs can portray readers regarding fluency. The context influenced the way MPLs and MPGs were processed, and should therefore be considered in the MPGs and MPLs processing architecture.