Projects per year
Understanding word-level emotion in terms of both category and intensity has always been considered an essential step in addressing text emotion classification tasks. Existing studies have mainly adopted the categorical lexicons that are tagged by predefined emotion taxonomies to link affective words with discrete emotions. However, in these lexicons, emotion tags are restricted to a specific set of basic emotions. Moreover, the emotional intensity is ignored, making these methods less flexible and less informative. This paper proposes a novel method to generate a word-level emotion distribution (WED) vector by incorporating domain knowledge and dimensional lexicon. The proposed method can link a word with more generic and fine-grained emotion taxonomies with quantitatively computed intensities. We propose two schemas to utilize the WED vector implicitly and explicitly to facilitate classification. The implicit approach implements a rule-based conversion strategy to augment the information in the label space. The explicit approach exploits WED as an emotional word embedding to enhance the sentiment feature. We conduct extensive experiments on seven multiclass datasets. The results indicate that both proposed schemas produce competitive results compared with the state-of-the-art baselines.
Bibliographical noteFunding Information:
The work in this research has been supported by the Faculty Research Fund ( 102041 ) and the Lam Woo Research Fund ( LWI20011 ) of Lingnan University, Hong Kong, the One-off Special Fund from Central and Faculty Fund in Support of Research from 2019/20 to 2021/22 ( MIT02/19-20 ), the Research Cluster Fund ( RG 78/2019-2020R ), the Interdisciplinary Research Scheme of the Dean’s Research Fund 2019-20 ( FLASS/DRF/IDS-2 ) of The Education University of Hong Kong, Hong Kong, and the Hong Kong Research Grants Council, Hong Kong under the Collaborative Research Fund ( C1031-18G ).
© 2021 Elsevier B.V.
- Emotion classification
- emotional embedding
- Emotional lexicon