In this study, we create an emotion lexicon for the Hindi language called Hindi EmotionNet. It can assign emotional affinity to words in IndoWordNet. This lexicon contains 3,839 emotion words, with 1,246 positive and 2,399 negative words. We also introduce ambiguous (217 words) and neutral (95 words) emotions to Hindi. Positive emotion words covered nine types of positive emotions, negative emotion words covered eleven types of negative emotions, ambiguous emotion words covered seven types of ambiguous emotions, and neutral emotion words covered two neutral emotions. The proposed Hindi EmotionNet was then applied to opinion classification and emotion classification. We introduce a centrality-based approach for emotion classification that uses degree, closeness, betweenness, and page rank as centrality measures. We also created a dataset of Hindi based on screenplays, stories, and blogs in the language. We translated emotion data from SemEval 2017 into Hindi for further comparison. The proposed approach delivered promising results on opinion and emotion classification, with an accuracy of 85.78% for the former and 75.91% for the latter.

中文翻译：

---

印地语情感网

在这项研究中，我们为印地语创建了一个名为 Hindi EmotionNet 的情感词典。它可以将情感亲和力分配给 IndoWordNet 中的单词。该词典包含 3,839 个情感词，其中 1,246 个积极词和 2,399 个消极词。我们还向印地语引入了模棱两可（217 个词）和中性（95 个词）的情绪。积极情绪词涵盖九种积极情绪，消极情绪词涵盖十一种消极情绪，暧昧情绪词涵盖七种暧昧情绪，中性情绪词涵盖两种中性情绪。然后将提出的印地语 EmotionNet 应用于意见分类和情绪分类。我们引入了一种基于中心度的情感分类方法，该方法使用度数、接近度、介数和页面排名作为中心度度量。我们还根据该语言的剧本、故事和博客创建了印地语数据集。我们将 SemEval 2017 中的情绪数据翻译成印地语以进行进一步比较。所提出的方法在意见和情感分类方面取得了可喜的成果，前者的准确率为 85.78%，后者的准确率为 75.91%。