Abstract
People's conduct and reactions are driven by their emotions. Online social media is becoming a great instrument for expressing emotions in written form. Paying attention to the context and the entire sentence help us to detect emotion from texts. However, this perspective inhibits us from noticing some emotional words or phrases in the text, particularly when the words express an emotion implicitly rather than explicitly. On the other hand, focusing only on the words and ignoring the context results in a distorted understanding of the sentence meaning and feeling. In this paper, we propose a framework that analyses text at both the sentence and word levels. We name it CEFER (Context and Emotion embedded Framework for Emotion Recognition). Our four approach facets are to extracting data by considering the entire sentence and each individual word simultaneously, as well as implicit and explicit emotions. The knowledge gained from these data not only mitigates the impact of flaws in the preceding approaches but also it strengthens the feature vector. We evaluate several feature spaces using BERT family and design the CEFER based on them. CEFER combines the emotional vector of each word, including explicit and implicit emotions, with the feature vector of each word based on context. CEFER performs better than the BERT family. The experimental results demonstrate that identifying implicit emotions are more challenging than detecting explicit emotions. CEFER, improves the accuracy of implicit emotion recognition. According to the results, CEFER perform 5% better than the BERT family in recognizing explicit emotions and 3% in implicit.
Abstract (translated)
URL
https://arxiv.org/abs/2209.13999
