Extending the EmotiNet Knowledge Base to Improve the Automatic Detection of Implicitly Expressed Emotions from Text

摘要

Sentiment analysis is one of the recent, highly dynamic fields in NaturalLanguage Processing. Most existing approaches are based on word-levelanalysis of texts and are mostly able to detect only explicit expressions ofsentiment. However, in many cases, emotions are not expressed by usingwords with an affective meaning (e.g. happy), but by describing real-lifesituations, which readers (based on their commonsense knowledge) detectas being related to a specic emotion. Given the challenges of detectingemotions from contexts in which no lexical clue is present, in this article wepresent a comparative analysis between the performance of well-establishedmethods for emotion detection (supervised and lexical knowledge-based) anda method we propose and extend, which is based on commonsense knowledgestored in the EmotiNet knowledge base. Our extensive evaluations showthat, in the context of this task, the approach based on EmotiNet is themost appropriate.