Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SenTopX: Benchmark for User Sentiment on Various Topics

Hina Qayyum, Muhammad Ikram, Benjamin Zhao, Ian Wood, Mohamad Ali Kaafar, Nicolas Kourtellis

TL;DR

SenTopX addresses the lack of longitudinal toxicity data on Twitter by assembling 143K users and 293M tweets from 2007–2021. It employs a contextualized topic model with BerTweet embeddings to extract 200 topics, categorizes them into eight groups, and computes per-user CPVs to form eight user groups. Toxicity is then analyzed across these groups using 16 Perspective API models, revealing distinct patterns such as elevated toxicity in the Everyday group and high toxicity across all metrics in Entertainment, with moderation implications. The dataset and methodology provide a robust framework for longitudinal toxicity research and platform moderation, with public availability for further exploration.

Abstract

Toxic sentiment analysis on Twitter (X) often focuses on specific topics and events such as politics and elections. Datasets of toxic users in such research are typically gathered through lexicon-based techniques, providing only a cross-sectional view. his approach has a tight confine for studying toxic user behavior and effective platform moderation. To identify users consistently spreading toxicity, a longitudinal analysis of their tweets is essential. However, such datasets currently do not exist. This study addresses this gap by collecting a longitudinal dataset from 143K Twitter users, covering the period from 2007 to 2021, amounting to a total of 293 million tweets. Using topic modeling, we extract all topics discussed by each user and categorize users into eight groups based on the predominant topic in their timelines. We then analyze the sentiments of each group using 16 toxic scores. Our research demonstrates that examining users longitudinally reveals a distinct perspective on their comprehensive personality traits and their overall impact on the platform. Our comprehensive dataset is accessible to researchers for additional analysis.

SenTopX: Benchmark for User Sentiment on Various Topics

TL;DR

SenTopX addresses the lack of longitudinal toxicity data on Twitter by assembling 143K users and 293M tweets from 2007–2021. It employs a contextualized topic model with BerTweet embeddings to extract 200 topics, categorizes them into eight groups, and computes per-user CPVs to form eight user groups. Toxicity is then analyzed across these groups using 16 Perspective API models, revealing distinct patterns such as elevated toxicity in the Everyday group and high toxicity across all metrics in Entertainment, with moderation implications. The dataset and methodology provide a robust framework for longitudinal toxicity research and platform moderation, with public availability for further exploration.

Abstract

Toxic sentiment analysis on Twitter (X) often focuses on specific topics and events such as politics and elections. Datasets of toxic users in such research are typically gathered through lexicon-based techniques, providing only a cross-sectional view. his approach has a tight confine for studying toxic user behavior and effective platform moderation. To identify users consistently spreading toxicity, a longitudinal analysis of their tweets is essential. However, such datasets currently do not exist. This study addresses this gap by collecting a longitudinal dataset from 143K Twitter users, covering the period from 2007 to 2021, amounting to a total of 293 million tweets. Using topic modeling, we extract all topics discussed by each user and categorize users into eight groups based on the predominant topic in their timelines. We then analyze the sentiments of each group using 16 toxic scores. Our research demonstrates that examining users longitudinally reveals a distinct perspective on their comprehensive personality traits and their overall impact on the platform. Our comprehensive dataset is accessible to researchers for additional analysis.
Paper Structure (16 sections, 1 equation, 1 figure, 1 table)

This paper contains 16 sections, 1 equation, 1 figure, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Dataset Collection
  4. Topic modeling
  5. Contextualized topic model
  6. Data preprocessing
  7. BERTweet textual embedding
  8. Model training
  9. Inferencing
  10. Hardware
  11. Interpretation of extracted topics and topic categories
  12. Topic categories
  13. Normalized global topic probability vector per user
  14. User groups based on engagement across 8 categories
  15. Toxicity analysis of user groups
  16. ...and 1 more sections

Figures (1)

  • Figure 1: Perspective API models scores; each plot represents mean score per user in each of 8 groups detailed in Section \ref{['sec:Thematic diversity']}, while red boxplots in each plot show overall scores.