SENCE-MEDIA

A real-time analytics program for detection of Hate speech on social media and analysing sentiments of news media articles.

Table of Contents

• SENCE-MEDIA
  • The Goal
  • Project Architechture
  • Services
  • APIS
  • How to get started

---

The Goal

The idea is to detect hate-speech against individual, community, organization, company on social media and use the data for a better visual representation for analytics.

Along with hate-speech, the project also focuses on sentiment analysis of news media articles about any of the above mentioned entity and present the resultant data in a dashboard.

Dashboard Demo

Since, it is a real-time pipeline, the data producer services run every day and the dashboard refreshes every 10 sec.

Project Architechture

• Data Producers : There are 2 producers, one producing Tweets and the other one that is responsible for sending news articles. The producer services send the messages via a Kafka topic.
• Data Consumers : Like producers, there are 2 consumers. One for tweets and another for news articles. On the consumer, the Expert.ai NLP API is called for classification and speech detection of the received messages. The classified data is then sent to the Elasticsearch indexes, that is used by Kibana for dashboarding.
• Apache Kafka : Has two topics, "topic-1" and "topic-2", each with 2 partition and 2 replicas.
• Elasticsearch & Kibana : Two separate indexes for storing news articles and tweets. Kibana access the data for visualization.

Services

• Producer : The producer services are named as send_news and send_tweets.
• Consumer : The consumer services are put as seperate services, name receive_news and receive_tweets.

APIS

• Twitter V2 API
• NewsAPI
• Expert.ai

How to get started

• Clone the repository

  $ git clone https://github.com/Dipankar-Medhi/sense_media.git
  
  $ cd sense_media
  

• Make a virtual environment and install the requirements

  $ python -m venv venv
  
  $ venv\Scripts\activate
  
  $ pip install -r requirements.txt
  

• Start the containers. Requires docker to be installed

  $ docker-compose up --build
  

• Once all the containers are set up and running well, start the services.

  • News articles service

    • Before starting the service, a Elasticsearch index must be created first.

    • An index can be create either by the dev tools at localhost:5601 or using the elasticsearch package.

    • There is a part of code commented out that helps to create an index.

    • Next, Start the receive news service.

      $ cd receive_news
      $ uvicorn main:main.py --reload
      

    • Next start the send news (producer) service.

      cd send_news
      $ uvicorn main:main.py --reload
      

    • We should be able to see live messages on discovery portal in Elasticsearch running at localhost:5601

  • Tweet service

    • Similar to News service, start with receive_tweets.
    • Then start the send_tweets service.