Coronavirus Twitter Sentiment Analysis

The sentiment analysis of Indians after the extension of lockdown announcements to be analyzed with the relevant #tags on twitter and build a predictive analytics model to understand the behavior of people if the lockdown is further extended. Also develop a dashboard with visualization of people reaction to the govt announcements on lockdown extension.

Objective

Twitter hosts abundant user-generated posts, which closely reflect the public’s reactions towards this pandemic with low latency.
Collecting and analyzing different demographic data helps compare and contrast that how different groups are affected in society.

Libraries

• Numpy :http://www.numpy.org/
• Pandas :http://pandas.pydata.org
• matplotlib :http://matplotlib.org/
• seaborn :https://seaborn.pydata.org
• Datetime :https://docs.python.org/3/library/datetime.html
• Statistics :https://docs.python.org/3/library/statistics.html
• textblob :https://textblob.readthedocs.io/en/dev/

Software Recommended

• Jupyter Notebook
• Anaconda installed with Python 3.8
• IBM Cloud, used IBM Watson Tone Analyzer service

Project Miletones

1. Data Gathering
2. Data Preprocessing
3. Analyzing Polarity and Subjectivity
4. Analyzing emotions using IBM Watson Tone Analyzer
5. Data Visualization
6. Creating Visualization Dashboard

Data Gathering
Tweets are scraped using web-scraping. Scraped data is stored in JSON format and then combining each tweet, converted into .csv file. CSV file then goes for data preprocessing.

Data Preprocessing

1. Remove Punctuations
2. Tokenizing Words
3. Stop words Removal
4. Stemming
5. Lemmatizing

Analyzing Polarity and Subjectivity
Each Tweet will be subdivided according to Polarity and Subjectivity: Polarity: Ranges from -1 to +1, -1 being most negative and +1 being most positive comment. Subjectivity: Ranges from 0 to 1, w.r.t its analytics. TextBlob has predefined functions and NLP models embedded in it, which makes this task efficient and smooth.

Analyzing tweet using IBM Watson Tone Analyzer
Analyzes the sentiment in the deeper way to classify and reveal the trend more comprehensively.
This technique uses linguistic analysis to detect joy, sadness, anger, fear, analytical, tentative and also confidence tones found in text.

Data Visualization
Technologies used:

1. Matplotlib
2. Plotly
3. Tableau
4. Seaborn

Creating Visualization Dashboard
Stack used:

1. HTML
2. CSS
3. JavaScript
4. Tableau
5. Bootstrap

Steps:

1. Run the data cleaning notebook to pre-process and clean the dataset.
2. Run the IBM Tone Analyser to call the IBM TOne Analyser API and analyse the data to categorise the tweets into 7 emotions.
3. Run the Sentiment Dataset Creation notebook to segregate the different emotion tweets and generate separate datasets for each emotion.
4. The Emotion Variation notebook is then used to analyse the variation of emotions in the tweet datsset.
5. Finally the polarity and Subjectivity notebook is run to generate the polarity vs subjectivity graph.
6. Use the different emotions datasets, import them in Tableau and generate dashboards and different graphs.
7. Embed the graphs in the website.