5.5 Conclusions
The application of the PLE method helped answer a common question to students: “What is the purpose of what I am studying?”. Answering this question helped awaken in students an interest in aiming to learn more and it also helped them to see how it will be possible to apply knowledge acquired in college in the professional life. The accomplishment of this case study indicates that the use of (PLE) favored the development of the reasoning in the future professionals, being important to mention that the students were determined to look for new cases to try to solve them or even to create new projects as alternatives for existing cases. Participating students were very keen to learn, observed that it is possible to apply the concepts acquired in the classroom in their professional life and developed the ability to work in teams respecting divergent opinions.
References
1. MINISTÉRIO DA EDUCAÇÃO. Ministério da Educação: Parâmetros Curricu-lares Nacionais. Brasília: MEC/SEF, 1997. Available in: <http://portal.mec.gov.br /seb/arquivos/pdf/livro01.pdf>. Accessed on: 1 May 2019.
2. TAVARES, S. R. T.; CAMPOS, L. C.; CAMPOS, B. C. O. Análise das abordagens PBL e PLE na Educação em Engenharia com base na Taxonomia de Bloom e no Ciclo de Aprendizagem de Kolb.Revista Eletrônica Engenharia Viva, Goiânia, p.
37-46, 2014. Available in: <https://www.revistas.ufg.br/revviva/article/view/292 54>. Accessed on: 1 May 2019.
CHAPTER 6
Automatic Correction of Discursive Question: An Approach to Contexts with Limited Language
Sandrerley Ramos Pires*, Dulcinéia Gonçalves Ferreira Pires␄, and Tobias Gonçalves Pires␞
*Escola de Engenharia Elétrica, Mecânica e de Computação, Universidade Federal de Goiás, Goiânia, Brazil E-mail: [email protected]
␄Centro Universitário de Patos de Minas, Patos de Minas, Minas Gerais, Brazil E-mail: [email protected]
␞Instituto Federal de Goiás, Campus Senador Canedo, Senador Canedo, Brazil E-mail: [email protected]
␟Goiânia, Brazil E-mail: [email protected]
Abstract:Distance learning platforms usually apply objective assessments to students, proposing and automatically correcting questions. A bank of issues enables this process. Although with recognized didactic value, this process is not ideal. Essay questions are too important because, in addition to testing students’ knowledge, they develop their ability to produce textual content. The automatic correction of discursive issues is an unresolved problem by the Na-tural Language Processing (NLP) area, but there are several approaches with partial solutions. The language offers us several words and syntactic construc-tions. Thus, there are several ways to express an answer. This fact is the main obstacle to the automatic correction of subjective questions. This work starts from the hypotheses that a human being invests decades in the process of refinement of his communication capacity, giving him large capacity of text production. Thus, it is likely that in adolescence their capacity for textual ex-pression is still limited. Proposing models to construct an answer can limit still more the variations of the textual expression. This work proposes an es-say question corrector that acts in a restricted language context. The main restrictions applied in the context were the application aimed at adolescents with a mean age of 15 years, the limitation of the programmatic content that
87
involves the issues and a template to structure the answers. These facts already reduce significantly the possibilities of textual expression. Other constraints imposed were (a) the creation of questions “what is”, “when it arose”, “what is the purpose of”; (b) establish the beginning of each answer; (c) the treatment of nominated entities; (d) the elimination of words without semantic value in the sentence and (e) the standardization of more frequent vocabularies used by the group of students. The correction process has the following steps: (a) the determination of the grammatical class of all words in the text; (b) the separation of sentences into clauses or noun phrase and (c) the creation of logical predicates representing the semantics contained in the analyzed phrase.
It was observed that for each model of the phrase syntactic structure, one can derive a set of logical predicates that express the idea contained therein. With eighteen syntactic rules it was possible to map the knowledge contained in sixteen responses of the class. Some semantic problems were found inspiring the research continuity. The measure of similarity between the logical proposi-tions that represent the answer given by the student and the proposiproposi-tions that represent the feedback of the question defines the student grade. The initial results demonstrate that the proposed algorithm can correct the discursive issues of a second-year high school class. In some cases, it is necessary to know the semantic of words to mapping the logical predicates, however, in the most of them, the phrase structure is enough for such mapping. The small number of necessary rules for interpretation of the studied questions shows the viability of the proposed approach.
Keywords:Automatic Correction, Discursive Question, Delimitation of Lan-guage Context, NLP.
6.1 Background
The automated process of learning assessment is performed by systems that use a bank of objective questions. The system presents to the evaluated some questions and make the automatic correction by the direct comparison with the feedback. The use of objective questions is not characterized as a problem1, but its exclusive use in an evaluation process is rather a problem2. It is important that automatic assessment also contain discursive question.
This kind of questions are important because, in addition to testing students’
knowledge, they develop their ability to produce textual content. However, in a discursive answer the student has several possibilities of using the natural language to present the answer.
This freedom to write the answer is the main problem that Natural Language Processing area3faces to make the automatic correcting of discursive questions.
It is a complex problem which complete solution is not yet available, although some approaches present good results in the content interpretations task.
The correction of a discursive questions is a problem of automatic interpre-tation of textual content4and of representation of interpreted knowledge. In addition, the proposed approach here involves the comparison of knowledge bases to measure how much a discursive question is correct.
This work argues that with insertion of some constraints in the natural language context, the possible ways to write an answer will be reduced. With this reducing, the system can automatically correct the answer possibilities remain.