Title of article :
Stance detection on social media, case study: Persian sentences using deep learning architecture
Author/Authors :
Mohammadi ، S. M. Department of Mathematics and Computer Science - Islamic Azad University, Arak Branch , Farzi ، S. Department of Computer - K. N. Toosi University of Technology , Alavi ، S. M. Department of Mathematics and Computer Science - Islamic Azad University, Arak Branch , Heydary Joonaghany ، Gh. Department of Mathematics and Computer Science - Islamic Azad University, Arak Branch
Abstract :
In the paper, we need to identify the stance of persian language in social networks. While the data set for detecting the stance with persian content have applications. Therefore, with the aim of accurately identifying the stance in the post and extracting the stances of persian language, the user expressed a post that relation to one or more target entities a new method for the first time. Hybrid LSTM-CNN architecture was used and, unlike previous researches, rotational learning rate was used, and a new the method for processing data before entering the network is presented to improve the results, which can stance persian in the network. Identifying the social in addition, to solve the problems related to the lack of data, the BERT model was investigated to detect the persian stance. On the based on the results obtained, tagged data was collected, and after many surveys and numerous meetings. Taged data analyzed from the telegram social network in the field of business sport for a limited time frame, and show how the presented model has achieved higher accuracy than Competitors. At the end of the training course, the proposed model improves results by 11 .3% in terms of accuracy.
Keywords :
Deep Learning , Persian Sentence , Stance Detection , Long Short Term Memory Networks
Journal title :
Scientia Iranica(Transactions D: Computer Science and Electrical Engineering)
Journal title :
Scientia Iranica(Transactions D: Computer Science and Electrical Engineering)