Do Angry People Have Poor Grammar?
summary
Uses a 1% sample of every public reddit comment (filtered to >=5 words, ~319,000 comments) and applies Python NLP tools: tokenization via NLTK, sentiment scoring with VADER, simple grammar/style checks via a linter, and regression/Pearson tests. Finds overall sentiment is mostly neutral and Pearson correlations between negativity and lint-flagged style issues are near zero with very small p-values, so anger and poor grammar show no meaningful statistical dependence at this scale.
AI-written summary · about
chapters (12)
- 0:00Sentiment and grammar research
- 3:35Reddit comment data
- 6:27Grammar with parsing
- 12:58Writing style with linters
- 17:16Sentiment with VADER
- 22:21Style versus sentiment
- 25:53Results and conclusions
- 27:41Future work and limits
- 28:03Q&A: Subreddit variations
- 34:46Q&A: Short comments bias
- 36:13Q&A: Top-rated political comments
- 39:12Q&A: Multilingual sentiment
related talks
53mA Machine Learning mood-o-meter
39mSentiment Analysis, its techniques and applications
43mGetting Sentimental: exploration of iOS sentiment analysis
01h 05mSocial Text, Sentiment, and Tone Analysis
26mReview Analysis: an Approach to Leveraging User-Generated...
45mAspect Based Sentiment Analysis in 20+ Languages
01h 23mBeyond Sentiment Emotion Mining with Python and machine learning
39mThe art and science of making computers understand sentiment and emotion