7. With SENTIMENT ANALYSIS we can
tackle the first block of questions:
7
What did people think?
What did they like and what didn’t they like?
What were people most excited about?
8. 8
SENTIMENT ANALYSIS
vaderSentiment is:
- rule and lexicon-based
- easy to use
- assigns polarity and intensity
- handles social media usage & emojis
- handles negation, i.e.
“VADER is not smart, handsome, nor
funny.”
16. 16
SENTIMENT ANALYSIS
“Steph Curry got outscored by Fred VanVleet & Kyle
Lowry at home in an elimination game in the NBA Finals.”
{'neg': 0.0,
'neu': 1.0,
'pos': 0.0,
‘compound': 0.0}
23. 23
With TOPIC CLUSTERING we can
tackle the second block of questions:
What were most people saying (i.e. what were the
trends in conversation)?
24. 24
TOPIC CLUSTERING
BERT:
- Bidirectional Encoder Representations
from Transformer
- Generates Word Embeddings
Word Embeddings
- Multi-dimensional numerical
representations of the context in which a
word is found