Measuring bias among text data using NLP methods

YouTube became one of the main digital education mediums during the pandemic.Previously, it was found that there exists a positive bias towards males in face-tofaceeducation. For instance, in one study, it was found that the attitude of theteachers while grading changed when they received the exam papers together withthe student names. Teachers graded male students more generously compared totheir females i.e. exhibited positive discrimination. In another study, the responsivenessof the professors to their emails was measured. The researchers concludedthat the professors are the most responsive when the email belongs to a white malestudent. Gender bias found in face-to-face education is also reflected in digital educationplatforms. In a study, the experimenters created a set of videos for science,technology, engineering mathematics (STEM), and the remaining fields of education(non-STEM). They gauged the gender bias and suggested that there is a bias towardsmales in both STEM and non-STEM videos although the the degree betweenthe two differs from each other. The goal of this study is to explore the reasonbehind this situation as well as understand the impact of the COVID-19 pandemicon the video and comment characteristics shared on the digital platforms. Our dataincludes 19867 educational video details collected from YouTube as well as their toprankedcomments. These videos were made by different narrators from January 2007to March 2021, and they were grouped based on STEM, and non-STEM queries.We focus on finding important evidence related to gender bias by working on thedifferences in the video and comment details such as the number of likes or views they get, the polarity of the comments, and the rank of the most common words. Inthis regard, we used a large variety of data preprocessing, statistical analyses, andsentiment analysis te