@inproceedings{baez-santamaria-etal-2024-contextualized,
    title = "Contextualized Graph Representations for Generating Counter-Narratives against Hate Speech",
    author = "Baez Santamaria, Selene  and
      Gomez Adorno, Helena  and
      Markov, Ilia",
    editor = "Al-Onaizan, Yaser  and
      Bansal, Mohit  and
      Chen, Yun-Nung",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2024",
    month = nov,
    year = "2024",
    address = "Miami, Florida, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/jlcl-multiple-ingestion/2024.findings-emnlp.450/",
    doi = "10.18653/v1/2024.findings-emnlp.450",
    pages = "7664--7674",
    abstract = "Hate speech (HS) is a widely acknowledged societal problem with potentially grave effects on vulnerable individuals and minority groups. Developing counter-narratives (CNs) that confront biases and stereotypes driving hateful narratives is considered an impactful strategy. Current automatic methods focus on isolated utterances to detect and react to hateful content online, often omitting the conversational context where HS naturally occurs. In this work, we explore strategies for the incorporation of conversational history for CN generation, comparing text and graphical representations with varying degrees of context. Overall, automatic and human evaluations show that 1) contextualized representations are comparable to those of isolated utterances, and 2) models based on graph representations outperform text representations, thus opening new research directions for future work."
}