Vol.30, No.2, 2021, pp.495-511, doi:10.32604/iasc.2021.018896
Research on Viewpoint Extraction in Microblog
  • Yabin Xu1,2,*, Shujuan Chen2, Xiaowei Xu3
1 Beijing Key Laboratory of Internet Culture and Digital Dissemination Research, Beijing, 100101, China
2 Computer School, Beijing Information Science and Technology University, Beijing 100101, China
3 Department of Information Science, University of Arkansas at Little Rock, Little Rock, 72204, USA
* Corresponding Author: Yabin Xu. Email:
Received 25 March 2021; Accepted 29 April 2021; Issue published 11 August 2021
In order to quickly get the viewpoint of key opinion leaders(KOL) on public events, a method of opinion mining in Weibo is put forward. Firstly, according to the characteristics of Weibo language, the non-viewpoint sentence recognition rule is formulated, and some non-viewpoint sentence is eliminated accordingly. Secondly, based on the constructed FastText-XGBoost viewpoint sentence recognition model, the second classification is carried out to identify the opinion sentence according to the dominant and recessive features of Weibo. Finally, the group of evaluation object and evaluation word is extracted from the opinion sentence, according to our proposed multi-task learning BiLSTM-CRFs model. In design, the “BIO” tagging mode is adopted. The sequence tagging of evaluation object and evaluation word based on LSTM-CRFs is conducted as the main task, and the loss function of the main task is optimized by the part of speech tagging based on BiLSTM-CRFs. The experiment result shows that the view recognition model based on FastText-XGBoost has obvious advantages over other recognition models in classification efficiency and accuracy, and the results of the MTL-BiLSTM-CRFs mining is more accurate and the model is more applicable.
Viewpoint extraction; opinion sentence recognition; recessive features; multitask learning; sequence tagging
Cite This Article
Xu, Y., Chen, S., Xu, X. (2021). Research on Viewpoint Extraction in Microblog. Intelligent Automation & Soft Computing, 30(2), 495–511.
This work is licensed under a Creative Commons Attribution 4.0 International License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.