Social Media Post Detection using Natural Language Processing and Machine Learning Classifiers

The increase in exposure to various viewpoints information on common social media necessitates the detection of bias in online discussions. The objective of this study is to implement an automated mechanism to categorize the text bias of social media information through Natural Language Processing and machine learning algorithms. Using a dataset of social media communications identified as partisan or neutral, this study illustrated different ways of approaching text bias. To identify which models for detecting bias in social media postings are the most effective the last models evaluated are several machine learning classifiers (classifiers that learned from pre-processed data using feature extraction) including SVM, Random Forest, Logistic Regression and Extra Trees. Perform necessary preprocessing steps including tokenization, stop word removal and lemmatization to clean and documented the text information. These images can assist in putting giant, over-utilized phrases into perspective of partisan vs neutral messaging so as to build insight into some terms occurring in reference to text talk. Comparison of individual model performance in regards to measures of accuracy, precision, recalls and F1-score which will form comparison benchmarks. SVM demonstrates the highest performances of accuracy, precision and the value of recall amongst the all-status classifiers. In general, this study denotes the possibilities of automating the text bias detection in social media with the help of machine learning technologies that can be related to the improvement of a more rapid and efficient analysis of the text contents. The system offers the tools, which may potentially increase the media literacy level, encourage critical thinking, and assist in detecting biased information in the world of fast digital interaction.