Covid19TwitterSentiment-Analysis

In this project we had worked for Covid19 Twitter Sentiment Analysis.
The dataset which we used to train our data consisted of tweet along with the original user sentiment at time of tweet. From a survey it also consisted of time taken for writing tweet and 1-10 scale for each individual feeling used

Prerequisite:

Present Work

1. Knowledge of Django (This project used Django as framework)
2. Knowledge of RNN (Bidirectional LSTM) Algorithm
3. Knowledge of NLP

Further modifications:

4. Twitter developer id (for live sentiment analysis)
5. Knowledge of IBM Cloud (for deployment)

Dataset link : https://arxiv.org/pdf/2004.04225.pdf

Code part

1. For installing libraries:

conda install -c anaconda pillow
conda install -c conda-forge matplotlib
conda install -c anaconda seaborn
conda install tensorflow

For running server:

python manage.py runserver

Frontend Backend integration

~ using function
~ connections made and call in url.py
~ functions in views.py
~ url activate in html file

html(activation of function)-> url(cheking function) -> views.py(checking function definition)

Content

Frontend content :

                  Main->Static Folder-> Trial Analysis2
                  Main-> Templates->Html 

Django files:

               Main-> Twwet_Dashboard-> models.py->(forms)
                                       urls.py->(paths for webpages)
                                       views.py-> (working functions of backend)
                                       Management-> commands-> bluemix_init.py(for hosting) 

Jupyter notebooks:

                   Jupyter_notebooks->Multifeeling_value.ipynb( For multiple sentences/lines output)
                                    ->twwet_me.ipynb ( For single line output )

for preprocessing

                   tokenizer-> tokenizer_SAVED_OOV.pickle   (tokenize (oov token)) 
                   (padding fn)[in jupyter notebook]

For dataset:

             dataset->hell.csv

For weights:

              weights-> multi_traget_feeling.hdf5 (multiple feelings for a particular text as outcome as bar graph in dashboard)
                     -> first_model_feeling1longonly_NEWMAIN.hdf5(single feeling as a summary of multiline text)

Some Documents:

Documentation link : https://drive.google.com/file/d/1E4MIv14svusdBCJkg6E3bNFCsnhkysSj/view?usp=sharing

PPT Link: https://drive.google.com/file/d/1fyJgoPZ6R57VXBwPeVX8mYqay2lOtYBV/view?usp=sharing

Video Link : https://drive.google.com/file/d/1LYOSQZQHyf8ZVZgsoek9iCeLQJCBde99/view?usp=sharing

Sample text to be test: https://drive.google.com/file/d/1D_1HkI-xMGVw1PotbrnCsSswF8Byvbis/view?usp=sharing

All Files link : https://drive.google.com/drive/folders/1PzMCkXa3VQy1cj36E2ulXMNXDTC2R6jk?usp=sharing

Research Papers for documentation

1. http://www.cs.columbia.edu/~julia/papers/Agarwaletal11.pdf
   
2. https://towardsdatascience.com/twitter-sentiment-analysis-based-on-news-topics-during-covid-19-c3d738005b55
   
3. https://arxiv.org/pdf/2003.05004
   
4. https://www.jmir.org/2020/4/e19016
   
5. https://arxiv.org/pdf/2003.10359
   
6. https://www.researchgate.net/profile/Kia_Jahanbin2/publication/339770709_Using_twitter_and_web_news_mining_to_predict_COVID-19_outbreak/links/5e84d4db4585150839b508b7/Using-twitter-and-web-news-mining-to-predict-COVID-19-outbreak.pdf
   
7. https://arxiv.org/pdf/2003.12309
   
8. https://arxiv.org/pdf/2004.04225
   
9. https://towardsdatascience.com/how-are-americans-reacting-to-covid-19-700eb4d5b597?source=rss----7f60cf5620c9---4