Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unveiling Social Media Comments with a Novel Named Entity Recognition System for Identity Groups

Andrés Carvallo, Tamara Quiroga, Carlos Aspillaga, Marcelo Mendoza

TL;DR

The paper addresses the challenge of hate speech on social media by moving beyond sentence-level detection to a Named Entity Recognition system that identifies and links attacks to specific identity groups. It achieves this by aligning two datasets— HateNorm (span-level entity mentions) and Jigsaw Toxicity (sentence-level labels)—to train a transformer-based NER that tags identity-group mentions across Religion, Ethnicity, Sexual Orientation, and Gender. The authors validate the approach with a case study on ABC News articles and Facebook comments, reporting competitive per-class F1-scores and uncovering patterns such as pronounced gender and sexual orientation mentions and low inter-category intersections, along with some false positives. Overall, the work offers a granular, scalable tool for analyzing hate speech and enables more nuanced moderation and intersectional analysis across social media content.

Abstract

While civilized users employ social media to stay informed and discuss daily occurrences, haters perceive these platforms as fertile ground for attacking groups and individuals. The prevailing approach to counter this phenomenon involves detecting such attacks by identifying toxic language. Effective platform measures aim to report haters and block their network access. In this context, employing hate speech detection methods aids in identifying these attacks amidst vast volumes of text, which are impossible for humans to analyze manually. In our study, we expand upon the usual hate speech detection methods, typically based on text classifiers, to develop a Named Entity Recognition (NER) System for Identity Groups. To achieve this, we created a dataset that allows extending a conventional NER to recognize identity groups. Consequently, our tool not only detects whether a sentence contains an attack but also tags the sentence tokens corresponding to the mentioned group. Results indicate that the model performs competitively in identifying groups with an average f1-score of 0.75, outperforming in identifying ethnicity attack spans with an f1-score of 0.80 compared to other identity groups. Moreover, the tool shows an outstanding generalization capability to minority classes concerning sexual orientation and gender, achieving an f1-score of 0.77 and 0.72, respectively. We tested the utility of our tool in a case study on social media, annotating and comparing comments from Facebook related to news mentioning identity groups. The case study reveals differences in the types of attacks recorded, effectively detecting named entities related to the categories of the analyzed news articles. Entities are accurately tagged within their categories, with a negligible error rate for inter-category tagging.

Unveiling Social Media Comments with a Novel Named Entity Recognition System for Identity Groups

TL;DR

The paper addresses the challenge of hate speech on social media by moving beyond sentence-level detection to a Named Entity Recognition system that identifies and links attacks to specific identity groups. It achieves this by aligning two datasets— HateNorm (span-level entity mentions) and Jigsaw Toxicity (sentence-level labels)—to train a transformer-based NER that tags identity-group mentions across Religion, Ethnicity, Sexual Orientation, and Gender. The authors validate the approach with a case study on ABC News articles and Facebook comments, reporting competitive per-class F1-scores and uncovering patterns such as pronounced gender and sexual orientation mentions and low inter-category intersections, along with some false positives. Overall, the work offers a granular, scalable tool for analyzing hate speech and enables more nuanced moderation and intersectional analysis across social media content.

Abstract

While civilized users employ social media to stay informed and discuss daily occurrences, haters perceive these platforms as fertile ground for attacking groups and individuals. The prevailing approach to counter this phenomenon involves detecting such attacks by identifying toxic language. Effective platform measures aim to report haters and block their network access. In this context, employing hate speech detection methods aids in identifying these attacks amidst vast volumes of text, which are impossible for humans to analyze manually. In our study, we expand upon the usual hate speech detection methods, typically based on text classifiers, to develop a Named Entity Recognition (NER) System for Identity Groups. To achieve this, we created a dataset that allows extending a conventional NER to recognize identity groups. Consequently, our tool not only detects whether a sentence contains an attack but also tags the sentence tokens corresponding to the mentioned group. Results indicate that the model performs competitively in identifying groups with an average f1-score of 0.75, outperforming in identifying ethnicity attack spans with an f1-score of 0.80 compared to other identity groups. Moreover, the tool shows an outstanding generalization capability to minority classes concerning sexual orientation and gender, achieving an f1-score of 0.77 and 0.72, respectively. We tested the utility of our tool in a case study on social media, annotating and comparing comments from Facebook related to news mentioning identity groups. The case study reveals differences in the types of attacks recorded, effectively detecting named entities related to the categories of the analyzed news articles. Entities are accurately tagged within their categories, with a negligible error rate for inter-category tagging.
Paper Structure (7 sections, 5 figures, 6 tables)

This paper contains 7 sections, 5 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. NER for identity Groups
  4. Case study
  5. Methodology and main findings
  6. Inspection of relevant cases
  7. Conclusions

Figures (5)

  • Figure 1: Pipeline Diagram for Classification, NER, and Identification of Text Spans Attacking Entity Groups
  • Figure 2: Comparative NER Annotations Identifying Attacked Entities and Their Text Spans
  • Figure 3: News (1) showing interactions between users. Our tool detects some gender and sexual orientation attacks during the conversation.
  • Figure 4: News (7) showing interactions between users. Our tool detects some attacks and mentions regarding racism and ethnicities during the conversation.
  • Figure 5: News (15) showing interactions between users. Our tool detects some attacks and mentions regarding sexual orientation during the conversation.