Aspect term extraction and snake driving training-based optimization- enabled deep learning for Facebook sentiment analysis

S Kotagiri and M Owk and T Priyanka and AM Sowjanya, KNOWLEDGE AND INFORMATION SYSTEMS, 67, 9437-9466 (2025).

DOI: 10.1007/s10115-025-02510-6

In this modern world, social media plays a vital role in communicating with each other across the globe. Sentiment analysis plays an essential role on social media platforms in tracking online conversations among themselves and with competitors in real time. They also gain measurable insights about how positively or negatively they are viewed by the users. In this research, a Facebook sentiment analysis approach named snake driving training-based optimization-enabled random multimodal deep learning (SDTBO_RMDL) is developed. Initially, the acquisition of Facebook review from the BuzzFeed News dataset is done. Then, to break down the sentence into tokens, BERT tokenization is carried out. After that, the aspect terms in the sentences are identified using the ATE. Next, the extraction of relevant features is done on the raw data using feature extraction. Then, sentiment analysis classification is performed using RMDL, where the network parameters are optimally tuned using the proposed SDTBO. Here, the SDTBO is the integration of driving training- based optimization and snake optimizer. Finally, the classified output comes from the RMDL. Thus, the Facebook reviews are categorized as positive as well as negative. The performance of SDTBO_RMDL is of values 0.948, 0.974, and 0.961 for precision, recall, and F1-score, respectively.

Return to Publications page