Sentiment analysis

From wiki: open source software tools deploy machine learning, statistics, and natural language processing techniques to automate sentiment analysis on large collections of texts, including web pages, online news, internet discussion groups, online reviews, web blogs, and social media.

Evaluation: 70-80% is OK.

Projects

• CyberEmotions consortium
• SentiStrength -- estimates the strength of positive and negative sentiment in short texts, even for informal language. Article "Sentiment strength detection in short informal text": http://onlinelibrary.wiley.com/doi/10.1002/asi.21416/full
• WordNet is a large lexical database of English. Nouns, verbs, adjectives and adverbs are grouped into sets of cognitive synonyms (synsets). Not specifically sentiment oriented, but has been used to help derive sentiment related information
• SentiWordNet (based on WordNet) -- assign each word of WordNet three sentiment scores: positivity, negativity, objectivity.
• Stanford Sentiment Analysis (web demo: http://nlp.stanford.edu:8080/sentiment/rntnDemo.html)

Tested

1. Stanford Sentiment Analysis (download: http://nlp.stanford.edu/sentiment/code.html)

   Written in Java. 5 classes: very negative, negative, neutral, positive, and very positive.

   It just works.

   To test: java -cp "*" -mx5g edu.stanford.nlp.sentiment.SentimentPipeline -file foo.txt

2. SentiStrength (http://sentistrength.wlv.ac.uk/, download Java version here: https://www.dropbox.com/s/st8eg4dy98lxn2y/SentiStrength.zip)

   Estimates the strength of positive and negative sentiment in short texts, even for informal language. Article "Sentiment strength detection in short informal text": http://onlinelibrary.wiley.com/doi/10.1002/asi.21416/full This system was successfully used in several research projects.

   Main version is a Windows executable, but Java version also exists and works.

   Many different classifiers: dual (two scales, for positive and negative emotions), binary (just 1 or -1), trinary (-1, 0, 1).

   Quick start: java -jar SentiStrength.jar sentidata ./SentiStrength_data/ text "i+don't+hate+you"