The art and science of making computers understand sentiment and emotion
summary
Claims concrete outcomes: a working emoji-suggestion pipeline trained on ~900,000 tweets (140,000 with emoji), 137 frequent emojis, train 82k/val 2k/test 2k, where adding an emoji matrix or emoji clusters raises accuracy/F-score (logistic 0.19 → 0.23, deep stack F1 0.21 → 0.27). Shows that simple bag-of-words often hits ~80% baseline that is hard to beat. Demonstrates sarcasm detection using incongruity signals and word embeddings plus eye-tracking yields ~70% accuracy. Uses CNN→LSTM→feedforward pipelines, spectral clustering and standard classifiers.
AI-written summary · about
chapters (11)
- 0:00Introduction to sentiment
- 2:02The NLP framework
- 5:33Defining sentiment analysis
- 8:00Dimensions of sentiment
- 10:47Methodologies and accuracy
- 14:53Emoji processing
- 19:01Emoji matrix implementation
- 25:12Deep learning architectures
- 30:13Sarcasm detection
- 34:02Cognition-based detection
- 36:00Conclusions and Q&A
related talks
37mMultilingual sentiment analysis & emotion detection based on social media data
32mDecoding Emotions from Art to Memes
39mSentiment Analysis, its techniques and applications
39mAI + JavaScript Rocks
01h 23mBeyond Sentiment Emotion Mining with Python and machine learning
59mTechnical Workshop: Sarcastically Speaking: Unlocking Multi-modal Sentiment Analysis With NLP and Facial Expressions
49mAffective Computing – Bringing Humans and Machines Closer Through Emotions
15mEmpathic dialogue system based on emotions extracted from tweets