TwitterSentiment is a Keyword based Twitter Sentiment Analyzer. It uses tweets based on the input provided by the user to generate a Rudimentry Sentiment Report.
https://twittersa-hyo.oa.r.appspot.com/
The Sentiment Model is trained on sentiment140 dataset. https://www.kaggle.com/kazanova/sentiment140
The dataset contains:
- 800k Positive Tweets
- 800k Negative Tweets
The 1.4 Million tweets are preprocessed using the following steps:
- LowerCase the letters
- Replacing the urls with "URL"
- Removing Usernames (@donaldtrump)
- Removing Special Characters i.e: Non-Alpha Numeric Characters
- Removing Stopwords
- Word lemmatization
The preprocessed tweets are then vectorized using Tf-idf. The vectorized tweets are used as a input for the Support Vector Machine Classifer.
⚠️ Docker Linux is needed for this!
docker pull realdexter/twitter_sentiment:v1
docker run -p local_port:8501 realdexter/twitter_sentiment:v1
local_port is the Port you want to map to the exposed port of the container.
Visit https://localhost:8501
⚠️ Model files not included in git repo.
git clone https://github.com/AsadAliDD/TwitterSentiment
pip3 install -r requirements.txt
streamlit run SentimentAnalysis.py
- Python3
- Streamlit
- Nltk
- Sklearn
- Pandas
- Numpy
- GetOldTweets3
- Plotly