Discovery of Emotion Implicit Causes in Products Based on Commonsense Reasoning
Lists
Only you see this
Abstract
This paper focuses on the task of product emotion cause analysis which aims to find essential causes of certain emotions from product reviews. Current works only study the explicit causes which are some spans in the given text. However, some crucial and useful causes may be expressed vaguely. They may not mention but can be inferred from the text semantics. They can capture the deeper reasons to explain some unknown phenomena, which can well support the applications like market research and product optimization. To address this problem, we in this paper propose a new task of Emotion Implicit Cause Discovery (EICD). We develop a novel method that can deduce the implicit causes based on the contexts and commonsense knowledge. Our method first retrieves related knowledge from the large language model to construct reasoning graphs for the emotions and potential causes. We then encode the structural knowledge in the graph and infer the implicit cause by deductive reasoning. To evaluate our method, we construct a large dataset called EICDset, based on Amazon product reviews. Experiments on it demonstrate the effectiveness of our model.
BibTeX
@inproceedings{guo2023discovery,
author = {Qiutong Guo and Jianxing Yu and Yufeng Zhang and Haowei Jiang and Wei Liu and Jian Yin},
title = {{Discovery of Emotion Implicit Causes in Products Based on Commonsense Reasoning}},
booktitle = {ADMA},
year = {2023},
doi = {10.1007/978-3-031-46661-8\_19},
}Loading related content...