Exploration of Automatic Spam/Ham Message Classifier Using NLP

Abstract

Text classification holds great significance for analyzing textual information in one or more predetermined categories. Text classification tools allows to arrange all different kinds of texts, emails, court records, advertisements, databases, and other materials. This saves time in making informed decisions based on accurate information. A text classifier is software functionality designed to segment unresolved text into a number of preordained groups within a given text. Such software is frequently incorporated into various applications, including sentiment analysis, online content moderation, spam detection in emails, and language detection. Classifying a text is an essential but complex undertaking, as it requires a level of language comprehension akin to that of humans. Consequently, text classification has emerged as a subject of ongoing research interest. This study’s primary goal is to investigate how cutting-edge natural language processing techniques are used to automatically detect spam messages. The proposed model will examine two distinct datasets to construct and categorize the disaster tweets. One dataset was obtained from the NUS SMS Corpus (NSC), while the other was specifically generated for this study using Grumbletext. When a spam email turns up, the system will inevitably identify the category using procedures like GloVe and BERT. The methodology, Naive Bayes DTM and XGBoost functionalities are used. The system’s performance will be assessed through metrics such as precision, recall, f-score, FNR, FPR.

Topics & Concepts