An Exploratory Study to Determine the Sentiment of the Indian OTT Web Series Audience by Applying Opinion Mining
Opinion Mining + Sentiment Classification for the Top 10 Indian Web Series (namely in 6 Genres - Action, Thriller, Drama, Comedy, Romance, Horror) using Python as my Intern Work @ IIM Ranchi.
In this project, we aim to perform Sentiment Analysis of user reviews of Top 10 Indian Web Series in different Genres (namely in 6 Genres - Action, Thriller, Drama, Comedy, Romance, Horror) shown on various OTT (over-the-top) platforms or media services (Amazon Prime Video, Netflix, ZEE5, Disney + Hotstar, Voot, ALTBalaji, SonyLiv etc). Data used in this project(roughly 50k+) was collected by performing Web Scraping on review platforms/websites like IMDb, Amazon, Google Reviews etc.
Here, we performed a fair amount of research on how sentiments are expressed in genres such as in online reviews and critics’ articles, how sentiments are expressed given the informal language and message-length constraints of user reviewing and then had drawn out useful inferences from them by identifying positive, negative, and neutral reviews based on their polarity and subjectivity.For reviews conveying both a positive and negative sentiment, whichever had the stronger sentiment was chosen. Additionally, created some data visualizations (bag of words, word clouds, plots etc) using the high frequency found sentiment keywords.
- A thorough study of existing approaches and techniques in field of sentiment analysis.
- Collection / Extraction of relevant data from review platforms/websites like IMDb , Amazon , Google Reviews etc. with the help of Web Scraping techniques using Python or Web Scraping Google Extensions ( like Instant Data Scraper ).
- Pre-processing of data collected so that it can be fit for mining. Basically, the unstructured raw data undergoes various steps of preprocessing and cleaning which makes it more machine sensible than its previous form.
- Finally, the reviews were analyzed, visualized and classified with respect to sentiment classes based on the TextBlob provided Polarity and Subjectivity , such as Positive, Negative and Neutral.
Opinion Mining, Sentiment Analysis, Machine Learning(ML), Natural Language Processing(NLP), Python, Text Mining, Data Preprocessing, Exploratory Data Analysis(EDA), Data Visualization, Web Scraping.