Место издания:Springer, Cham A.V., Rozaliev, V.L., Orlova, Y.A., Zaboleeva-Zotova
Первая страница:23
Последняя страница:30
Аннотация:This article is devoted to the method of recognition of sarcasm in the text in order to improve the accuracy of sentiment analysis. As a domain area for the experiment, the task of determination the sentiment level of a text that describes the appearance of a person was chosen. At first, references to the personality and elements that describes appearance from the text written in a natural language are detected using the method of latent semantic analysis. Frame model of knowledge representation was chosen as the information structure of the human appearance. The next step is to evaluate the attitude to the person in the text using the pre-labeled sentiment dictionary. At this stage, the method of recognizing sarcastic sentences that contains a description of the appearance is used. The sentiment level should be re-evaluated in the slots of the frame instance. The F-measure was used as an efficiency criterion for correct sarcasm recognition. At the end of the article, the results of experiment are provided.