Sentiment Analysis for Hungarian language

This is open-source sentiment analysis tool for Hungarian language, written in Python. The application has a REST API for easier access, and also accessible via Docker's container technology.

##Installation: Docker container installation is suggested. ###1. Native way

• Prerequisite: linux Operation System
  • Ubuntu, tested for only 14.04 LTS
  • Debian, RHEL, CentOS done, but not yet tested for more distributions
• Launch ./install.sh as root privileged user and accept every step of installation.
• Takes about 20-30 minutes

###2. Docker

• Install Docker for you Operation System: (https://docs.docker.com/engine/installation)
• Docker repo: (https://hub.docker.com/r/dhuszti/sentanalysis/)
• Download container: docker pull dhuszti/sentanalysis
• Run container: docker run -d -it --privileged=true --net=host -e LANG=C.UTF-8 --name=sentanalysishun dhuszti/sentanalysis

##Usage: ###1. Launch application with:

• Native: python $HOME/SentimentAnalysisHUN-master/src/Application.py
• Docker: docker exec -it sentanalysishun python $HOME/SentimentAnalysisHUN-master/src/Application.py

###2. Usage of REST API:

• Based on a HTTP POST request.
• Receive an url (substitute to example): http://ip_addr_of_docker:5000/sentiment
• Please use sentence tag for adding input like this {'sentence': '<write your input here>'} to enter your input.
• Tool has two different HTTP POST requests:
  • /sentiment: for overall score
  • /sentiment_verbose: for more detailed scores
• Example usage on Linux/Mac with console curl: curl -i -H "Content-Type: application/json" -X POST -d '{"sentence": "Budapest az egyik legszebb város."}' http://<ip_of_your_machine>:5000/sentiment
• For Windows use a REST client like https://github.com/wiztools/rest-client

##Sources Following external open-source tools were applied, some of their installer files are collected at /resources folder. All of their rights are owned by their creators.

1. NLP tools - HunToken, HunPoS, Ocamorph & HunMorph, Polyglot NER
2. OpinHUBank sentiment corpus
3. Python Flask REST API - http://flask.pocoo.org/
4. Python packages - NumPy, SciPy, Sklearn, NLTK