Unsupervised emotion detection for Twitter with sarcasm detection

Social media has become a common avenue for transmission of information. There has been a rising trend in research on sentimental analysis and opinion mining on Twitter in the recent years due to the popularity of Twitter. The aim of these research is to develop ways to extract sentiments or opinion...

Full description

Saved in:
Bibliographic Details
Main Author: Sim, Jun Shen
Other Authors: Ke Yi Ping, Kelly
Format: Final Year Project
Language:English
Published: 2017
Subjects:
Online Access:http://hdl.handle.net/10356/70172
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Social media has become a common avenue for transmission of information. There has been a rising trend in research on sentimental analysis and opinion mining on Twitter in the recent years due to the popularity of Twitter. The aim of these research is to develop ways to extract sentiments or opinions of the public, which are beneficial in applications such as business and government intelligence. Many methodologies and approaches used for sentiment analysis and opinion mining on Twitter often faced difficulties in classifying tweets that are sarcastic in nature. Sarcasm is a special communication method that uses words that means opposite to what the author is trying to convey. The words used in a sentence may be positive in nature but the underlying emotion that was conveyed was a negative one. In the report, I propose a sentimental analysis model to incorporate a sarcasm detector into an existing sentiment analysis method to enhance the performance of the sentiment classification of a tweet. The sarcasm detector is based on explicit sarcastic labels found in the hashtags of the tweets. These sarcastic tweets are identified and removed from the test data. A self-generated lexicon approach was used to create a polarity dictionary which was then used to calculate and classify the remaining test data based on the polarity of tweets. The results show that the proposed method performed better than the original method when identifying both positive and negative.