Twitter Sentiment Analyzer

This application is a demonstration/proof of concept of sentiment analysis of a series/stream of tweets.

• Analysis is done by using Stanford Core NLP library and a trained model (Stanford Core NLP Model)
• Application makes use of few Akka Framework's libraries such as Akka streams and Akka HTTP to achieve the result

Application processes data in the following order:

1. It reads data from a source file in the root of the project containing basic tweet info (tweet id, URL, auth, etc.)
2. Extracts each tweet id, url and other info from each line and construct a basic tweet object
3. Using Twitter API, fetches a tweet by id
4. Runs tweet's text through the analyzer and gets a score
5. Constructs a tweet object with sentiment score
6. Writes data to a csv file in the root of the project

Note

• Data file contains many deleted tweets, when using those tweet ids if API call return error or failure, those tweets are dropped from stream processing
• NLP model Jar file is in /lib directory

Application makes use of Flows in Akka stream, demonstrating possibility of substituting Source or Sink with different sources such as a message bus (Queue, Stream or Database).

All flows are in Processor.scala (start method)

Source of input file: http://nlp.uned.es/replab2013/replab2013-dataset.tar.gz via https://github.com/awesomedata/awesome-public-datasets

Important

If any errors occurs from Stanford Core NLP complaining about NLP models missing, please download stanford-corenlp-4.2.2-models.jar from https://stanfordnlp.github.io/CoreNLP/download.html and add it to /lib directory.

How to Run

Run Main.scala using sbt:

sbt run