Online-Sentiment-Analysis-on-Twitter-Streams

Master Thesis "An Empirical Study of Online Sentiment Analysis on Twitter Streams " offered by DIMA, TU Berlin

• Author: Huilin Wu
• Advisor: Dr. Shuhao Zhang

Motivation

• Most existing studies regarding Sentiment Analysis are based on offline batch-based learning mechanisms. Meanwhile, many stream processing systems have been proposed, but they are not specifically designed for online learning tasks, such as online Sentiment Analysis. As a result, it still remains an open and challenging question of how to efficiently perform Sentiment Analysis for real-time streaming data, e.g., ongoing Twitter Streams.
• The goal of this thesis is to empirically evaluate various online algorithms for Sentiment Analysis on Twitter Streams by implementing them on DSPS (Data Stream Processing System) for practical application.

Environment Requirement

1. Flink v1.12
2. Scala v2.11
3. Python 3.7
4. Java 8
5. Kafka 2.13
6. Redis server v4.0.9

DataSource

1.6 million labeled Tweets: Sentiment140

Quick Guide

1. Pyflink_demo

• Tweets_clean_demo: Demo for batch-based Tweets preprocessing for Sentiment Analysis on FLink
• Streaming_demo_Sentiment_Analysis: Demo for stream-based Tweets preprocessing & online Sentiment Analysis model on Flink

2. OSA_algorithms

• Algorithms of incremental Sentiment Analysis

3. Python_small_job

• Python file of various developing note