Creating emoji lexica from unsupervised sentiment analysis of their descriptions
作者：
Highlights：
• Method to create emoji sentiment lexicon using unsupervised SA and Emojipedia.
• Unsupervised SA strategy based on semantic dependencies with propagation.
• Lexicon variants created considering sentiment distribution of messages with emojis.
• Lexica compare favorably well with other ones obtained manually or with CLDR names.
• Approach and variants applied to the Spanish, English and Spanish+English datasets.
摘要
•Method to create emoji sentiment lexicon using unsupervised SA and Emojipedia.•Unsupervised SA strategy based on semantic dependencies with propagation.•Lexicon variants created considering sentiment distribution of messages with emojis.•Lexica compare favorably well with other ones obtained manually or with CLDR names.•Approach and variants applied to the Spanish, English and Spanish+English datasets.
论文关键词：Emoji analysis,Sentiment analysis,Opinion mining,Nlp,Artificial intelligence
论文评审过程：Received 28 September 2017, Revised 25 February 2018, Accepted 26 February 2018, Available online 8 March 2018, Version of Record 20 March 2018.
论文官网地址：https://doi.org/10.1016/j.eswa.2018.02.043
