Sentiment Symposium Tutorial

Sentiment Analysis Symposium, San Francisco, November 8-9, 2011

Instructor: Christopher Potts (Stanford Linguistics)

Overview

1. Overview: goals, plan, and applications2
2. Sentiment in language and cognition13
3. Text preparation:
   1. Tokenization 23
   2. Stemming34
   3. Advanced linguistic structure40
4. Sentiment lexicons52
5. Classifier models for sentiment78
6. Vector space models95
7. Context-dependency and social relationships107
8. Sentiment summarization122
9. Bibliography133

Publicly-available tutorial data and implementations

• Basic, extensible Python sentiment tokenizer with random-tweet tokenizing function
• Word distributions:
  • Multi-corpus word distributions (zip archive)
  • Multi-linguial, multi-corpus word distributions
  • POS-tagged word distributions from IMDB (zip archive)
  • Multi-corpus statistics on WordNet adjectives and adverbs
  • Heavily documented R code for creating lexicons and scales from data like the above
• Large Movie Review dataset
• Simple WordNet sense propagation in Python/NLTK
• WordNet score propagation in Python/NLTK; example lexicon; documentation
• Cosine score propagation algorithm in Python