Nonstandard Emoticon Sentiment Classification

Abygail McMillan, University of North Georgia

Description/Abstract

Emoticon classification in sentiment analysis has typically followed that of sentiment classification of all text: falling under either positive, negative, or neutral, with minimal declared effect on the text. However, previous research has indicated that introducing an “in between” step, that of classifying emoticons under the categories of either intensification, negation, or only sentiment prior to evaluating the sentiment of the text as a whole, and giving the emoticons more weight during final analysis of the text, may make final sentiment analysis of the whole text more accurate. In this presentation we explore this system and its overall accuracy, using a manually annotated corpus of English Twitter posts (“tweets”) and the Natural Language Toolkit platform as a basis.

 
Mar 25th, 10:15 AM Mar 25th, 11:30 AM

Nonstandard Emoticon Sentiment Classification

Nesbitt 3100

Emoticon classification in sentiment analysis has typically followed that of sentiment classification of all text: falling under either positive, negative, or neutral, with minimal declared effect on the text. However, previous research has indicated that introducing an “in between” step, that of classifying emoticons under the categories of either intensification, negation, or only sentiment prior to evaluating the sentiment of the text as a whole, and giving the emoticons more weight during final analysis of the text, may make final sentiment analysis of the whole text more accurate. In this presentation we explore this system and its overall accuracy, using a manually annotated corpus of English Twitter posts (“tweets”) and the Natural Language Toolkit platform as a basis.