Abstract:

The increasing attention of researchers is being drawn to automated personality recognition (APR) due to the advancements in intelligent technologies, such as deep learning. APR is one of the approaches of personality computing to auto-assess personality. Research in the area of education, such as education data mining (EDM) and learning analytics (LA), typically necessitates the use of students’ private information and costly sensors that are challenging to acquire in practice. This research focuses on gathering data and performing an analysis of the APR in an actual university teaching setting.

Initially, data was collected from discussion-based and teaching-based courses, respectively. The collected data encompasses several forms, including vocal recordings, handwritten samples, and other class-related data such as scores. The collection process of these data does not involve student privacy, nor use expensive sensors, and can serve as a research basis for the generalization of APR in the field of education.

Furthermore, we performed a single-model and a multi-model automatic recognition study on the collected data, followed by data analysis. The single-model automatic speech recognition employs deep neural networks (DNNs) and achieves good experimental results with a low mean squared error (MSE), which is a value used to evaluate the performance of model. Multi-model tests were undertaken to analyze the APR and weight of each modality's data.

In conclusion, the data collected in the current university teaching setting may also be used for basic APR. This implies that APR in the area of education will no longer be restricted to study alone but can progress towards practical application.

Keywords:

Education, automatic personality recognition, multi-model, data analysis.