Poem Emotion Recognition Corpus (PERC)

Even though there is available of the lexicon in emotion analysis, to identify emotion from poems had to rely on limited emotion lexicons. Since those lexicons are not created for poems, and it is not concentrated on poetic features. This paper presents a text corpus PERC(Poem Emotion Recognition Corpus) comprising a set of poems and features for emotion recognition from poems. Emotion classi cation is based on 'Navarasa,' described in 'Natyasastra.' Navarasa consists of nine primary emotions such as Love, Sad, Anger, Hate, Fear, Surprise, Courage, Joy, and Peace. Although there are many text corpus for emotion recognition, we do not know of a text corpus for poems based on nine emotions. The corpus created is from an exhaustive collection of poems of Indian poets for the period 1850-2016. The novelty of this work is the creation of a corpus using poems mined from the web and evaluated by human experts.

尽管情感分析领域已存在可用的情感词典，但从诗歌文本中识别情感仍需依赖有限的情感词典。由于这些词典并非针对诗歌领域专门构建，且未聚焦诗歌文本特征，相关研究存在一定局限。本文提出了诗歌情感识别语料库（Poem Emotion Recognition Corpus，简称PERC），该语料库包含多首诗歌及用于诗歌情感识别的特征集。该语料库的情感分类依据《舞论》（Natyasastra）中记载的九情（Navarasa）体系，九情包含九种核心情感，分别为爱、悲伤、愤怒、憎恨、恐惧、惊讶、勇敢、喜悦与平和。尽管现有诸多面向通用情感识别的文本语料库，但目前尚无基于九种情感体系构建的诗歌专用语料库。本次构建的语料库全面收录了1850年至2016年间印度诗人的诗歌作品。本研究的创新之处在于，其构建的语料库数据源自网络爬取的诗歌，并经过了人类专家的审核与标注。