Skip to main content

Tools for collecting , processing and analyzing twitter data

Project description

Twitter Toolbox

A suite of tools for collecting, pre-processing, analyzing and sentiment-scoring twitter data. Install:

pip install twitter-nlp-toolkit

While the package is still under active development, the following functionality is expected to be stable:

Listener

twitter_nlp_toolkit.twitter_listener is the listener module, which can be used to monitor Twitter and stream tweets to disk in .json format.

keywords = ["python"]
stream = twitter_listener.TwitterStreamListener(**credentials)
stream.collect_from_stream(max_tweets=10,output_json_name="python_tweets.json", target_words=keywords)

"keywords" uses the Twitter API. Documentation and tips for setting up smart keyword queries can be found here: https://developer.twitter.com/en/docs/tweets/rules-and-filtering/overview/standard-operators

"credentials" contains your Twitter API key, which can be obtained for free here: https://developer.twitter.com/en/docs/basics/developer-portal/overview

The module also contains a parser to convert the .json-formatted tweets into .csv for easy use (ie, with Pandas).

parser = tweet_json_parser.json_parser()
parser.stream_json_file(json_file_name="python_tweets.json",output_file_name="parsed_python_tweets.csv")

Bulk Downloader

twitter_nlp_toolkit.twitter_REST_downloader is the bulk download module, which can be used to collect the last 200 (or so) tweets from a single user.

downloader = twitter_REST_downloader.bulk_downloader(**credentials)
downloader.get_tweets_csv_for_this_user("@nytimes","nyt_tweet_output.csv")

twitter_nlp_toolkit.tweet_sentiment_classifier is the sentiment analysis module, which can be used to classify the sentiment of tweets.

Classifier = tweet_sentiment_classifier.SentimentAnalyzer()
Classifier.load_small_ensemble()
Classifier.predict(['I am happy', 'I am sad', 'I am cheerful', 'I am mad']) # will return [1, 0, 1, 0]

Currently only two ensembles are provided: the small ensemble, which uses bag-of-words logistic regression model and two long short-term memory neural networks, and the large ensemble, which uses the bog-of-words model, two larger LSTM networks, and a Google BERT model. The large ensemble is more accurate (and expected to become much more accurate), but is extremely resource intensive and as such, isn't recommended for processing large numbers of tweets unless you have a powerful GPU.

These ensembles were trained primarily on the Sent140 dataset: http://help.sentiment140.com/ and primarily tested against the US Airlines dataset previously hosted on Crowdflower.com.

Please see the jupyter notebook (.ipynb) files at the root directory for further demonstrations of working code.

Advanced Use

If you have domain-specific training data, you can refine the ensembles:

Classifier.refine(train_x, train_y)
Classifier.save_models()
# To reload your saved models, you can run Classifier.load_models()

Other advanced use, such as building your own models, is possible but is not currently recommended as the models are still in development. Further documentation will be added once development stabilizes.

Bugs, issues, contributions, and feature requests

The developers are always open to feature requests, bugs reports, pull requests, and new opportunities to collaborate. Don't hesitate to reach out with questions, beedback, or requests.

Developers:

  • Moe Antar (@Moe520)

    • Twitter json parser and formatter
    • Twitter bulk downloader
    • Twitter Listener, in collaboration with Dr. Mirko Miorelli (https://github.com/mirkomiorelli)
    • File downloader
    • General maintenance and deployment procedures
  • Eric Schibli (@eschibli)

    • Data pre-processing pipeline
    • Natural language processing algorithms
    • Bag-of-words, LSTM neural networks, and BERT model assembly and training
    • General model optimization

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

twitter_nlp_toolkit-0.1.4.tar.gz (20.1 kB view hashes)

Uploaded Source

Built Distribution

twitter_nlp_toolkit-0.1.4-py3-none-any.whl (25.8 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page