Vaccine sentiment analysis using BERT + NBSVM and geo-spatial approaches

• DOI: 10.1007/s11227-023-05319-8
• Publication Date: 2023-05-07T17:01:15Z
• Authors: Areeba Umair; Elio Masciari; Muhammad Habib Ullah
• Keywords: Sentiment Analysis

Since the spread of coronavirus flu in 2019 (hereafter referred to as COVID-19), millions people worldwide have been affected by pandemic, which has significantly impacted our habits various ways. In order eradicate disease, a great help came from unprecedentedly fast vaccines development along with strict preventive measures adoption like lockdown. Thus, world wide provisioning was crucial achieve maximum immunization population. However, vaccines, driven urge limiting pandemic caused skeptical reactions vast amount More specifically, people's hesitancy getting vaccinated an additional obstacle fighting COVID-19. To ameliorate this scenario, it is important understand sentiments about take proper actions better inform As matter fact, continuously update their feelings and on social media, thus analysis those opinions challenge for providing information avoid misinformation. detail, sentiment (Wankhade et al. Artif Intell Rev 55(7):5731-5780, 2022. 10.1007/s10462-022-10144-1) powerful technique natural language processing that enables identification classification (mainly) text data. It involves use machine learning algorithms other computational techniques analyze large volumes determine whether they express positive, negative or neutral sentiment. Sentiment widely used industries such marketing, customer service, healthcare, among others, gain actionable insights feedback, media posts, forms unstructured textual paper, Analysis will be elaborate reaction COVID-19 provide useful improve correct understanding usage possible advantages. framework leverages artificial intelligence (AI) methods proposed classifying tweets based polarity values. We analyzed Twitter data related after most appropriate pre-processing them. we identified word-cloud negative, words using tool tweets. After step, performed BERT + NBSVM model classify vaccines. The reason choosing combine bidirectional encoder representations transformers (BERT) Naive Bayes support vector (NBSVM ) can understood considering limitation BERT-based approaches, only leverage layers, resulting lower performance short texts ones analysis. Such ameliorated Support Vector Machine approaches are able higher took advantage both features define flexible goal vaccine identification. Moreover, enrich results spatial geo-coding, visualization, correlation suggest suitable vaccination centers users outcomes. principle, do not need implement distributed architecture run experiments available public massive. discuss high-performance if collected scales up dramatically. compared approach state-of-art comparing metrics Accuracy, Precision, Recall F-measure. outperformed alternative models achieving 73% accuracy, 71% precision, 88% recall F-measure positive while 74% respectively. These promising properly discussed next sections. lead any trending topic. case health-related topics could implementing health policies. availability findings user policymakers design strategies ad-hoc protocols according feelings, service. end, leveraged geospatial effective recommendations centers.