Personalisation of reading through use of automatic text classification and adaptation: reading tool providing different levels of text complexity directed to students in Brazil´s basic school system

• Grant number: 15/08632-1
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Duration: December 01, 2015 - November 30, 2017
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Danielle Gobbi Brants
• Grantee: Danielle Gobbi Brants
• Host Company: Guten Educação e Tecnologia Ltda. - EPP
• City: São Paulo
• Pesquisadores principais: Livia Aluisi Cucatto ; Nathan Siegle Hartmann

Abstract

The need for substantial improvements in Brazilian education, especially related to the initial years of schooling, is recognized by all. Despite advances in the early stages of literacy, there has been little progress in the mastery of skills that are prerequisites for a complete integration of the individual into society. According to INAF, calculated from 2001 to 2012, 26% of the population presented a high degree of literacy, while 21% presented rudimentary levels (able to understand explicit information in short sentences) while 47% would only reach the basic level of proficiency (can read longer sentences and make simple inferences). The functional illiterates totaled 27% (Illiterate and Rudimentary). Reading processes and reading comprehension should be natural: respect individuality of the student and at the same time provide motivation for such activity. Any problems in the early stages of literacy acquisition can lead to blockages or detachment. Given this scenario, to adapt texts to the reading skills of students is crucial for their progress. The lack of technological resources that help reading acquisition in Brazil, by adapting the text to appropriate levels of readability, is one of the factors that prevent students to become better readers in adulthood. The importance of reading and text comprehension has attracted researchers from many countries and various applications, for example, in the construction of teaching materials for second language teaching (Collins-Thompson, 2011; Graesser et al 2011; Sheehan, et al., 2013 LoPucki, 2014; Febres and Klaus, 2014; François, 2014; Mihai, et al 2014; Dell'Orletta; Aluísio et al, 2010; Short, 2014; among others). Research and existing tools for Portuguese, however, do not solve the demands of educators in the classroom: no automatic classifiers are able to discriminate levels of text complexity according to school years, using metrics from the various levels of language. Guten sees this gap as a relevant business opportunity with a significant potential to impact the quality of Portuguese reading in Brazil. Thus, this project aims to create a reading tool than classifies texts into 5 readability levels, directed to students aged 8-12 years old. Therefore, the creation of a text classifier and textual adaptation software is needed, considering the specifics of Portuguese language and Brazil´s Educational market. The construction of the above mentioned requires language, computer, educational and statistical knowledge, especially the resources already developed in PorSimples Project (Aluísio and Gasperin, 2010). The business model contemplated follows the pattern already initiated by the applicant company in its previous products: the commercial approach will focus on private schools and school systems (B2B) and, in the future, municipal and state districts (B2G). Our objective is to contribute to the development of computational resources for Brazilian Portuguese directed to Education and to the improvement of reading instruction and comprehension in the country. (AU)