An efficient CNN-LSTM model for sentiment detection in #BlackLivesMatter

Creative Commons License

Ankita A., Rani S., Bashir A. K., Alhudhaif A., Koundal D., GÜNDÜZ E. S.

Expert Systems with Applications, vol.193, 2022 (SCI-Expanded) identifier

  • Publication Type: Article / Article
  • Volume: 193
  • Publication Date: 2022
  • Doi Number: 10.1016/j.eswa.2021.116256
  • Journal Name: Expert Systems with Applications
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, Aerospace Database, Applied Science & Technology Source, Communication Abstracts, Computer & Applied Sciences, INSPEC, Metadex, Public Affairs Index, Civil Engineering Abstracts
  • Keywords: Convolution neural network, Deep learning, Long Short Term Memory, Sentiment analysis, Social media
  • Akdeniz University Affiliated: Yes


Imagining things without mixed emotions is next to impossible in today's scenario. Whether it is news or any online movement started on social media applications. One of the social media applications i.e Twitter started a movement known as #BlackLivesMatter. The people from all over the world participated showing mixed reactions, sentiments, and emotions such as trusting the movement, gave negative feedback, felt disgusting, showing anger, etc. In this study, a deep learning classifier Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) is used to detect the sentiments and emotions of the people based on the tweets of the two provinces of the USA (Minnesota and Washington D.C.). The proposed hybrid model is validated over Random Forest, Convolutional Neural Network, Long Short Term Memory, and Bidirectional Long Short Term Memory. It is really surprising to see the results as in both the provinces people showing interest as they are trusting the movement with 48% in Minnesota and 54% in Washington D.C. Our proposed model CNN-LSTM is 94% accurate in detecting the various sentiments based on the hyper-parameters such as epoch, filter size, pooling, activation function, dropout, stride, padding, and number of filters.