Personality Traits Detection in Bangla: A Benchmark Dataset with Comparative Performance Analysis of State-of-the-Art Methods

摘要

Nowadays, people are interested in using various online platforms to share their opinions and thoughts on various topics and issues. The informal user-generated contents in these platforms make it an important source for studying and modeling the personality of a person. Detecting and analyzing user personality plays an important role to design an effective recommendation system, Q/A system for customer care, employee assessment, and product promotions. Prior works on personality detection from user-generated text mostly conducted on the English language. However, there is no previous work and dataset available for automatic detection of user personality from Bangla text. In this paper, we bridge this research gap and present a benchmark Bangla personality traits detection dataset that consists of 3000 Bangla informal text collected from various online platforms. Besides, we present various baseline systems by exploiting state-of-the-art supervised classification methods and perform a comparative performance analysis that provides an important insight about this task. We believe this dataset might be beneficial for others for developing more effective models and we publicly release the dataset for future research purposes at the following link: https://git.io/JkW6V or use the expanded URL ^1.