Advanced search
Start date
Betweenand

An automated platform for evaluations and analysis of essays to develop students' writing

Grant number: 17/00655-8
Support type:Research Grants - Innovative Research in Small Business - PIPE
Duration: February 01, 2019 - January 31, 2021
Field of knowledge:Linguistics, Literature and Arts - Linguistics
Cooperation agreement: FINEP - PIPE/PAPPE Grant
Principal Investigator:Clayton Quandt Dick
Grantee:Clayton Quandt Dick
Company:Nota 1000 Serviços Educacionais Ltda - ME
CNAE: Desenvolvimento e licenciamento de programas de computador não-customizáveis
Atividades de apoio à educação
Atividades de ensino não especificadas anteriormente
City: São Paulo
Co-Principal Investigators:Carolina Siequeroli
Assoc. researchers:Amanda Pontes Rassi ; Fernando Antônio Asevedo Nóbrega ; Júlia Sales Paez Fernandez
Associated scholarship(s):19/14034-0 - An automated platform for evaluations and analysis of essays to develop students' writing, BP.TT
19/01272-0 - An automated platform for evaluation and analysis of essays to develop students' writing, BP.TT
19/01923-1 - An automated platform for evaluation and analysis of essays to develop students' writing, BP.TT
18/24999-0 - An automated platform for evaluation and analysis of essays to develop students' writing, BP.TT

Abstract

Developing writing skills of students - and of people in general - has always been a major challenge. To begin with, writing essays is naturally difficult because people have to rely on several competences at once; moreover, regular practical exercises are required in order for writing skills to evolve. Writing practice, however, is hardly effective unless it is supported by proper assessment/guidance. Structured and clear feedback on errors and successes as well as suggestions on how to improve are crucial to students' effective development. Nevertheless, teachers often lack the time, resources or qualifications to offer proper feedback and guidance. As an attempt to bridge this gap, Redação Nota 1000 was created in 2013 as an online platform for production and evaluation of essays, with a view to improving students' writing through practice. Although the main purpose of the platform, since the beginning, was to automate this process in order to aid teachers-reviewers in the analysis of essays, there were not enough resources, databases and knowledge, at that time, to achieve this goal. Therefore, we have decided to validate the model, its impact and its marketing potential as a first step, and then start developing automated solutions. Now that we have successfully completed the first phase of the project, our next goal is to create an online tool for automatic correction. Our aim is to assist students, teachers and reviewers, to an even greater extent, in the production, correction, evaluation and rewriting of essays according to the standards established by ENEM (Brazil's High School National Exam) and other college entrance exams. To achieve this goal, our work is underpinned by three principles: The first is Machine Learning, i.e., a set of methods, rules and procedures which enable computers to make predictions and decisions based on standards that had not been defined previously. These methods are widely disseminated and present in almost all fields of automation - from Internet search algorithms to autonomous vehicle driving control systems. Another advantage is that they can be easily hired and adapted - also on demand. The second principle is relative to NLP (Natural Language Processing) tools and applications, i.e. computational-linguistic resources and software which process human language automatically. For this purpose, we recruit the best specialists in the academic environment and the best tools for linguistic annotation, at various levels: morphological, syntactic, semantic and discursive. The third and last principle is the database itself - Big Data - which we have created, based on the analysis and correction of hundreds of thousands of essays in the past four years. The database was designed particularly for this purpose, and it uses a very high degree of detail and structure. In short, we apply state-of- the-art features of machine learning and NLP to our massive database so that a computer system can "learn" the standards and criteria for evaluation in an increasingly accurate manner. Next, we will develop a predictive model capable of anticipating an ever greater percentage of potential writing errors. We have the resources, the methodology and a qualified and experienced team to automate the text assessment process. This project has not only a great immediate impact on education in Brazil, but also the potential to open a new opportunity for research in this field worldwide. (AU)