Distantly Supervised Explainable Stance Detection via Chain-of-Thought Supervision
Investigating public attitudes on social media is crucial for opinion mining systems. Stance detection aims to predict the attitude towards a specific target expressed in a text. However, effective neural stance detectors require substantial training data, which are challenging to curate due to the...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2024-04-01
|
| Series: | Mathematics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2227-7390/12/7/1119 |
| _version_ | 1827286685458628608 |
|---|---|
| author | Daijun Ding Genan Dai Cheng Peng Xiaojiang Peng Bowen Zhang Hu Huang |
| author_facet | Daijun Ding Genan Dai Cheng Peng Xiaojiang Peng Bowen Zhang Hu Huang |
| author_sort | Daijun Ding |
| collection | DOAJ |
| description | Investigating public attitudes on social media is crucial for opinion mining systems. Stance detection aims to predict the attitude towards a specific target expressed in a text. However, effective neural stance detectors require substantial training data, which are challenging to curate due to the dynamic nature of social media. Moreover, deep neural networks (DNNs) lack explainability, rendering them unsuitable for scenarios requiring explanations. We propose a distantly supervised explainable stance detection framework (DS-ESD), comprising an instruction-based chain-of-thought (CoT) method, a generative network, and a transformer-based stance predictor. The CoT method employs prompt templates to extract stance detection explanations from a very large language model (VLLM). The generative network learns the input-explanation mapping, and a transformer-based stance classifier is trained with VLLM-annotated stance labels, implementing distant supervision. We propose a label rectification strategy to mitigate the impact of erroneous labels. Experiments on three benchmark datasets showed that our model outperformed the compared methods, validating its efficacy in stance detection tasks. This research contributes to the advancement of explainable stance detection frameworks, leveraging distant supervision and label rectification strategies to enhance performance and interpretability. |
| first_indexed | 2024-04-24T10:38:48Z |
| format | Article |
| id | doaj.art-249cc4fd6f89478dbd6434d699a9ab41 |
| institution | Directory Open Access Journal |
| issn | 2227-7390 |
| language | English |
| last_indexed | 2024-04-24T10:38:48Z |
| publishDate | 2024-04-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Mathematics |
| spelling | doaj.art-249cc4fd6f89478dbd6434d699a9ab412024-04-12T13:22:57ZengMDPI AGMathematics2227-73902024-04-01127111910.3390/math12071119Distantly Supervised Explainable Stance Detection via Chain-of-Thought SupervisionDaijun Ding0Genan Dai1Cheng Peng2Xiaojiang Peng3Bowen Zhang4Hu Huang5College of Applied Science, Shenzhen University, Shenzhen 518052, ChinaCollege of Big Data and Internet, Shenzhen Technology University, Shenzhen 518118, ChinaSchool of Computing, Zhongshan Institute, University of Electronic Science and Technology of China, Zhongshan 528402, ChinaCollege of Big Data and Internet, Shenzhen Technology University, Shenzhen 518118, ChinaCollege of Big Data and Internet, Shenzhen Technology University, Shenzhen 518118, ChinaShenzhen Graduate School, Peking University, Shenzhen 518055, ChinaInvestigating public attitudes on social media is crucial for opinion mining systems. Stance detection aims to predict the attitude towards a specific target expressed in a text. However, effective neural stance detectors require substantial training data, which are challenging to curate due to the dynamic nature of social media. Moreover, deep neural networks (DNNs) lack explainability, rendering them unsuitable for scenarios requiring explanations. We propose a distantly supervised explainable stance detection framework (DS-ESD), comprising an instruction-based chain-of-thought (CoT) method, a generative network, and a transformer-based stance predictor. The CoT method employs prompt templates to extract stance detection explanations from a very large language model (VLLM). The generative network learns the input-explanation mapping, and a transformer-based stance classifier is trained with VLLM-annotated stance labels, implementing distant supervision. We propose a label rectification strategy to mitigate the impact of erroneous labels. Experiments on three benchmark datasets showed that our model outperformed the compared methods, validating its efficacy in stance detection tasks. This research contributes to the advancement of explainable stance detection frameworks, leveraging distant supervision and label rectification strategies to enhance performance and interpretability.https://www.mdpi.com/2227-7390/12/7/1119stance detectionprompt-tuningchain-of-thought |
| spellingShingle | Daijun Ding Genan Dai Cheng Peng Xiaojiang Peng Bowen Zhang Hu Huang Distantly Supervised Explainable Stance Detection via Chain-of-Thought Supervision Mathematics stance detection prompt-tuning chain-of-thought |
| title | Distantly Supervised Explainable Stance Detection via Chain-of-Thought Supervision |
| title_full | Distantly Supervised Explainable Stance Detection via Chain-of-Thought Supervision |
| title_fullStr | Distantly Supervised Explainable Stance Detection via Chain-of-Thought Supervision |
| title_full_unstemmed | Distantly Supervised Explainable Stance Detection via Chain-of-Thought Supervision |
| title_short | Distantly Supervised Explainable Stance Detection via Chain-of-Thought Supervision |
| title_sort | distantly supervised explainable stance detection via chain of thought supervision |
| topic | stance detection prompt-tuning chain-of-thought |
| url | https://www.mdpi.com/2227-7390/12/7/1119 |
| work_keys_str_mv | AT daijunding distantlysupervisedexplainablestancedetectionviachainofthoughtsupervision AT genandai distantlysupervisedexplainablestancedetectionviachainofthoughtsupervision AT chengpeng distantlysupervisedexplainablestancedetectionviachainofthoughtsupervision AT xiaojiangpeng distantlysupervisedexplainablestancedetectionviachainofthoughtsupervision AT bowenzhang distantlysupervisedexplainablestancedetectionviachainofthoughtsupervision AT huhuang distantlysupervisedexplainablestancedetectionviachainofthoughtsupervision |