Summary of Leveraging Large Language Models and Topic Modeling For Toxicity Classification, by Haniyeh Ehsani Oskouie et al.
Leveraging Large Language Models and Topic Modeling for Toxicity Classification
by Haniyeh Ehsani Oskouie, Christina Chance, Claire Huang, Margaret Capetz, Elizabeth Eyeson, Majid Sarrafzadeh
First submitted to arxiv on: 26 Nov 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A recent paper explores the relationship between content moderation, toxicity classification, and annotator positionality. The authors investigate how fine-tuning BERTweet and HateBERT on specific topics affects their performance in detecting text toxicity. The results show that fine-tuned models achieve a higher F1 score compared to other prominent classification models like GPT-4, PerspectiveAPI, and RewireAPI. However, the study also reveals significant limitations of large language models in accurately detecting and interpreting text toxicity, highlighting the need for further improvements. The paper contributes to the ongoing discussion on content moderation and toxicity classification by shedding light on the impact of annotator positionality on model performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research looks at how we can make computers better at recognizing when something is mean or harmful online. The study shows that making these computer programs learn specific topics helps them do a better job detecting bad language than other approaches. However, the researchers also found that even with this improvement, these “smart” computers still struggle to accurately identify toxic text. This paper aims to understand how we can make our content moderation tools more effective and fair. |
Keywords
» Artificial intelligence » Classification » F1 score » Fine tuning » Gpt
