From 2f8f0c2db109de58528cdb765e9b59a8ed46c35b Mon Sep 17 00:00:00 2001 From: Kshitiz Tiwari Date: Fri, 8 Dec 2023 23:05:31 -0600 Subject: [PATCH] Update informations --- .../2009-10-01-paper-title-number-1.md | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) diff --git a/_publications/2009-10-01-paper-title-number-1.md b/_publications/2009-10-01-paper-title-number-1.md index ca6a157f7902..0e0c7deb68da 100644 --- a/_publications/2009-10-01-paper-title-number-1.md +++ b/_publications/2009-10-01-paper-title-number-1.md @@ -1,15 +1,15 @@ --- -title: "Paper Title Number 1" +title: "Robust Hate Speech Detection via Mitigating Spurious Correlations" collection: publications -permalink: /publication/2009-10-01-paper-title-number-1 -excerpt: 'This paper is about the number 1. The number 2 is left for future work.' -date: 2009-10-01 +permalink: /publication/robust-hate-speech-detection +excerpt: 'This paper is about removing the causal connection of spurious correlated words to develop a robust hate speech detection model.' +date: 2022-10-01 venue: 'Journal 1' -paperurl: 'http://academicpages.github.io/files/paper1.pdf' -citation: 'Your Name, You. (2009). "Paper Title Number 1." Journal 1. 1(1).' +paperurl: 'https://aclanthology.org/2022.aacl-short.7' +citation: [Robust Hate Speech Detection via Mitigating Spurious Correlations](https://aclanthology.org/2022.aacl-short.7) (Tiwari et al., AACL-IJCNLP 2022) --- -This paper is about the number 1. The number 2 is left for future work. +We develop a novel robust hate speech detection model that can defend against both wordand character-level adversarial attacks. We identify the essential factor that vanilla detection models are vulnerable to adversarial attacks is the spurious correlation between certain target words in the text and the prediction label. To mitigate such spurious correlation, we describe the process of hate speech detection by a causal graph. Then, we employ the causal strength to quantify the spurious correlation and formulate a regularized entropy loss function. We show that our method generalizes the backdoor adjustment technique in causal inference. -[Download paper here](http://academicpages.github.io/files/paper1.pdf) +[Read Paper Here](https://aclanthology.org/2022.aacl-short.7.pdf) -Recommended citation: Your Name, You. (2009). "Paper Title Number 1." Journal 1. 1(1). \ No newline at end of file +Recommended citation: [Robust Hate Speech Detection via Mitigating Spurious Correlations](https://aclanthology.org/2022.aacl-short.7) (Tiwari et al., AACL-IJCNLP 2022) \ No newline at end of file