Sunday, March 31, 2013

Online Tools for Sentiment Classification - Part I

Seeing Sentiment Analysis in action is a good way of getting a feel for what the tasks are about, and what techniques are in use today. In this post I survey some of the research interfaces available online which you can use on your favorite piece of text. 

Christopher Potts Tutorial - Text Scoring Demo
In this tutorial demo by Stanford Professor Christopher Potts, an engine that analyses text and classifies sentiment according to hits from different sentiment lexicons.

http://sentiment.christopherpotts.net/textscores/

His tutorial page also links to code demos in Python showcasing a number of techniques for Sentiment Analysis. 


TweetFeel

TweetFeel is a polished UI for quickly inspecting sentiment on twitter posts, but is not so much a research project and more a demo of a commercial product (there is not a lot of details of underlying techniques here). My tests revealed much dismay on the twittersphere at Ireland's squad recent results at the WC qualifiers :(

http://www.tweetfeel.com/#ireland


Sentimentor
Sentimentor is a project from Univ. of Brighton's James Spencer. The output of this tool is a visual display of how each term was scored with respect to sentiment orientation. 

The method and code behind this interface is available upon request from the author.



Sentiment140

Sentiment140 is a project from Stanford University students Alec GoRicha Bhayani, and Lei Huang, based on their work on using machine learning algorithms to classify twitter sentiment. (requires a twitter account)

SentiWordNet

SentiwordNet is a sentiment lexicon built from WordNet. It assigns sentiment orientation to words by expanding an initial seed set by inspecting term relationships and sentiment glosses. The UI allows querying for a term, giving back numeric scores for a tuple (positive, negative, objective)


The lexicon is described on this paper (see here for v1.0), and is also available for download.

Other Resources




No comments:

Post a Comment