Abstract:
Sentiment analysis has been a most imperative and very hot area in natural language processing. The aim of sentiment analysis is to examine and extort understanding from subjective information that is presented on internet. Sentiment Analysis is the automated interaction of deciding if content communicates a positive, negative, or impartial assessment on an item or theme. By utilizing supposition investigation, organizations do not need to spend unlimited hours labeling client information, for example, overview reactions, audits, uphold tickets, and web-based media remarks. To recognize extremity, this strategy characterizes a bunch of rules utilizing Natural Language Processing (NLP) methods (like tokenization, stemming, and parsing) close by physically made standards. The epidemic of COVID-19 disease has rathlessly affected economies worldwide including Pakistan. Foremost sufferers of COVID-19 epidemic are small, intermediate size ventures COVID-19 badly affected all type of businesses. The COVID-19 epidemic, with millions of consumers, is behind the massive increase in e-commerce purchases Quarantine surfing for online reviews of products, services and entertainment has risen exponentially. As we know that as number of online consumer's increases during COVID-19 and people who are ordering for online food are also increased. Already research showed five credits specific to eatery surveys-food, service, ambience, cost and context-this investigation proposes Cleanliness as the 6th attribute interesting to online audits. We have added sixth attribute to existing datasets. In our proposed method we have implemented three Machine Learning models on two datasets that includes Restaurant review dataset and Yelp reviews dataset. The datasets are standardized to be utilized by Al calculations and arranged utilizing characteristic language handling procedures like word tokenization, stemming and lemmatization. The yelp dataset is investigated utilizing Multinomial Naive Bayes, Bernoulli Naive Bayes and Logistic Regression. Multinomial Naive Bayes gives the best accuracy of 90.91%, precision 91% and recall 91%. Further in this study we implemented one Deep Learning Model Long Short Term Memory (LSTM) on yelp dataset LSTM is applied on problem of order dependence, sequence prediction. Our problem is based on NLP in association of which LSTM performs much better because of its approach towards complex task is very effective. We have many parameters, reviews and rating, but we focused on those parameters where sufficient task and research have been conducted previously and these parameters ate food, price, ambience, and context. However, there is another parameter ie. Cleanliness on the bases of which we can say that our model performed efficiently on such grounds. The obtained accuracy by giving different parameters for example: Greeting, Food, Taste etc were apprising but when the parameter of Cleanliness applied we found least crror and very impressive responses and when we tried to predict the same on the bases of testing parameters reviews the obtained results was 4 out of 5 persams gives considerable importance to this factor. Model behaves in such way that the prediction was considered to 3-4 range Despite of other high yield parameter Cleanliness reflected prominent and Impressive results that we also provide customer service in it and people consider these things a lot, Food Quality, ambience, cleanliness, customer services and pricing determine the worth of restaurant standard.
