Analysis of Student Feedback by Ranking the Polarities

TLDR: Text preprocessing techniques which includes tokenization, parts of speech (POS), sentence split, lemmatization, gender identification, true case, named entity recognition (NER), parse, conference graph, regular expression NER, and sentiment analysis are introduced to improve more accurate results and giving importance even to insignificant details in the text.

Abstract: Feedbacks in colleges and universities are often taken by means of online polls, OMR sheets, and so on. These methods require Internet access and are machine dependent. But feedbacks through SMS can be more efficient due to its flexibility and ease of usage. However, reliability of these text messages is a matter of concern in terms of accuracy, so we introduce the concept of text preprocessing techniques which includes tokenization, parts of speech (POS), sentence split, lemmatization, gender identification, true case, named entity recognition (NER), parse, conference graph, regular expression NER, and sentiment analysis to improve more accurate results and giving importance even to insignificant details in the text. Our experimental analysis on sentiment trees and ranking of feedbacks produces exact polarities to an extent. By this way, we can determine better feedback results that can be supplied to the faculty to enhance their teaching process.