Student project and comment system performing feature based opinionmining with machine learning

The use of online learning systems is becoming more and more common. With the Covid-19 pandemic, its importance has increased even more. The distribution of homework and projects and the recollection of these works by the instructor is an important step in the transfer of education to the online environment. The transfer of homework and projects to the online environment and the students' exchange of ideas about the work of their peers led to the accumulation of comment data about assignments and projects. By means of these comments, putting forward a system that can reveal the opinions about assignments and projects to the instructor on a feature-based basis will provide convenience to the instructor in terms of time and effort during the evaluation phase. In the study, it is aimed to create a system that performs feature-based idea mining that can support the instructor. Various preprocessing steps were applied to the comments obtained in this thesis, and feature detection was carried out linguistically and with machine learning methods. With the linguistic method, a success rate of 0.18 was achieved in the feature extraction phase. An F1 score of 0.81 was obtained with the Conditional Random Fields Algorithm. Machine learning models were trained to determine polarities through manually labeled data. An F1 score of 0.56 was obtained with K- Nearest Neighbor Algorithm, 0.59 with Decision Trees Algorithm, 0.606 with Support Vector Machines and 0.64 with Random Forest Classifier. Through the results, the Student Project and Comment System, in which the features are determined with the Conditional Random Fields and the polarities of the features are determined with the Random Forest Classifier, has been developed.