Longitudinal Sentiment Classification of Reddit Posts

TL;DR

This study develops a sentiment classifier for Reddit posts from Canadian university students during 2020-2023, demonstrating consistent performance across multiple datasets by fine-tuning sentiment thresholds.

Contribution

Introduces a sentiment classification method with threshold tuning that achieves consistent results across diverse university datasets.

Findings

Sentiment classifiers achieved high accuracy with threshold tuning.

Classification results were consistent across four university datasets.

The method effectively distinguishes positive and negative sentiments.

Abstract

We report results of a longitudinal sentiment classification of Reddit posts written by students of four major Canadian universities. We work with the texts of the posts, concentrating on the years 2020-2023. By finely tuning a sentiment threshold to a range of [-0.075,0.075], we successfully built classifiers proficient in categorizing post sentiments into positive and negative categories. Noticeably, our sentiment classification results are consistent across the four university data sets.