BACKGROUND
Happiness refers to the joyful and pleasant emotions that humans produce subjectively. It is the positive part of emotions, and it affects the quality of human life. Therefore, understanding human happiness is a meaningful task in sentiment analysis. We mainly discuss two facets (Agency/Sociality) of happiness in this study. Through analysis and research on happiness, we can expand on new concepts that define happiness and enrich our understanding of emotions.
OBJECTIVE
In this paper, we treated each happy moment as a sequence of short sentences, then proposed a short happiness detection model based on transfer learning to analyze the Agency and Sociality aspects of happiness.
METHODS
Happiness analysis is a novel and challenging research task. However, the current dataset in the field of happiness is small. To solve this problem，we utilized the unlabeled training set and transfer learning to train a semantically enhanced language model in the target domain. Then, the trained language model with domain characteristics was further combined with other deep learning models to obtain various models. Finally, we used the improved voting strategy to further improve the experimental results.
RESULTS
The proposed approach was evaluated on the public dataset. Experimental results showed that our approach significantly outperforms the baselines. When predicting the Agency aspect of happiness, our approach achieved an accuracy of 0.8574 and an F1 score of 0.90, repectively. When predicting Sociality, our approach achieved an accuracy of 0.928 and an F1 score of 0.9360, respectively.
CONCLUSIONS
Through the evaluation of the dataset, the comparison results demonstrated the effectiveness of our approach for happiness analysis. Experimental results confirmed that our method achieved state-of-the-art performance and transfer learning effectively improved happiness analysis.
Exploring Versatile Generative Language Model Via Parameter-Efficient Transfer Learning
