Beyond Sentiment Emotion Mining with Python and machine learning
summary
Builds a Python pipeline that replaces crude positive/negative scores with a nuanced vector of emotions using Inside Out–style labels, trained on the IMDb reviews plus the NRC emotion lexicon, and shows which approaches work and which don’t: traditional bag‑of‑words and a support vector machine hit ~65% accuracy (humans ~85%), some emotions classify well while others confuse. Demonstrates LSI/word‑vector tricks with gensim and a NetworkX graph model that maps authors/domains to emotion nodes and that the graph approach proves useful enough for production experiments.
AI-written summary · about
chapters (8)
- 0:00Sentiment analysis basics
- 3:14Limitations of standard sentiment
- 9:37Human emotions and Inside Out
- 13:11Nuanced emotion mining models
- 24:18Preprocessing text for ML
- 35:33Vectorizing with TF-IDF and LSA
- 1:08:46Graph-based emotion mining
- 1:22:00Q&A
related talks
39mSentiment Analysis, its techniques and applications
32mUnpaired Sentiment-to-Sentiment Translation: Using Reinforcement Learning
39mThe art and science of making computers understand sentiment and emotion
32mDecoding Emotions from Art to Memes
21mEmotion Estimation for Social Robotics
01h 07mMachine Learning and JavaScript. Unleashing the power of sentiment analysis
25mSentiment Analysis using Support Vector Machines in Ruby
53mA Machine Learning mood-o-meter