TSA

Twitter Sentiment Analyzer using Gensim's Word2Vec + NLTK + Keras.

Requirements

Miner

⚠️To be able to use miner.py you must have a Twitter developer account.

Your authentication details will need to be placed directly inside miner.py OR in a file named creds.py that will look like this:

consumer_key = 'abcd'
consumer_secret = 'efgh'
access_token = '4567'
access_token_secret = '1234'

Analyzer

To use TSA you will need to open your virtual env and install the required files in requirements.txt with pip3:

pip3 install -r requirements.txt

If you use Anaconda for example, you can follow these steps:

conda create -n TSA python=3.6

conda activate TSA

pip3 install -r requirements.txt

Usage

You can start the program by running ./analyzer.py

You will then be greeted with a menu:

 ████████ ███████  █████  
    ██    ██      ██   ██ 
    ██    ███████ ███████ 
    ██         ██ ██   ██ 
    ██    ███████ ██   ██                          

---Welcome to TSA---
What would you like to do?

    1) Mine tweets
    2) Validate a json file
    3) Analyze a corpus of tweets
    4) Find the most common words in a corpus of tweets
    5) Predict a string
    6) Quit

Enter your choice (1-6): 

You will be prompted to enter a number in the range 1-6.

• 1) Mine tweets will prompt you to enter your queries, miner.py will then look for tweets containing these terms and they will be put in a file called tweets.json. You will be able to interrupt the program by pressing Ctrl+C.

• 2) Validate a json file will prompt you to enter a JSON file name. The file will then be reformatted into a valid JSON file and you will then be able to use it with the analyzer.

• 3) Analyze a corpus of tweets will prompt you to enter a JSON file name and will ask you if you also want to visualize extra data. The file will then analyze the sentiments in your corpus, you will see a prediction percentage for positive and negative tweets. If you agreed to get extra data, you will also get 4 extra charts: one showing the word embedding space in Sentiment140, one for Zipf's law, one for the word frequency distribution and one for bigrams.

• 4) Find the most common words in a corpus of tweets will prompt you to enter the number of the most common words you want to see and a JSON file name for your corpus. You will then see the N most common words in your corpus.

• 5) Predict a string will prompt you to enter a string you'd like to predict the sentiment of.

• 6) Quit will quit the program.

Credits

This program has been built with:

• Python 3.6
• tweepy
• gensim
• tensorflow
• nltk
• scikit-learn

⚠️Python 3.6 is recommended to run this program.