An Assessment Study Of Public Perception About Vaccination On Social Media Using Sentiment Analysis And Topic Modeling Approach

Text research was often used to look at anti-vaccine attitudes and is particularly timely considering the rise of medical misinformation on social media.As the world becomes more connected,sharing one’s thoughts and views across social networking platforms is becoming more popular.This is why Twitter,the world’s largest microblogging website,is so famous.Well-known politicians,comedians,and trending entities use this forum to convey themselves in 140character messages.As a result,Twitter has become one of the most influential think platforms in the global network culture.This research uses topic modeling and sentiment analysis to extract information from tweets.Topic modeling aims to decide what people chat about,while sentiment analysis aims to determine what people think.This study uses 89,973 consumer tweets from Twitter to assess public mood and opinion on vaccination from 2006 to the 30th of November 2019.We used the open-source knowledge tool TWINT(Twitter Intelligence Tool)to capture tweets and used the Python VADER(Valence Aware Dictionary for sEntiment Reasoning)library to quantify each sentence’s polarity,article,and paragraph and the gensim library’s Latent Dirichlet Allocation feature to conduct sentiment analysis and topic modeling.Then quantify everyday public opinion toward vaccination using sentiment analysis.The most prominent regular conversation topics on Twitter using the keyword "vaccination" are discovered using topic modeling.We used gensim to construct an informative topic model focused on the Latent Dirichlet Allocation(LDA)algorithm,and we followed a standard workflow to do so.First,tokenize and clean the sentence by deleting emails,newline characters,and single quotations,then break the sentence into a list of terms and remove punctuations using gensim.Second,build a bigram and trigrams model to improve the execution speed,and then lemmatize each word to its root type,retaining only nouns,adjectives,verbs,and adverbs.Only these POS tags are kept because they are the ones that have the most significant impact on the interpretation of the sentences.To construct the LDA topic model,we create a corpus and a dictionary.Finally,each text is split into many parts.However,only one of the topics is usually influential.So,for each sentence,we extract the dominant subject and display the weight of the topic and the keywords in nicely formatted output.Regular Twitter conversation topics,discovered using unsupervised machine learning,were also a strong proxy for significant current events relevant to vaccination.