Sentiment Analysis

This project is my introduction to Sentiment analysis. Hopefully some cool stuff will result from this!

Exploring the VADER model:

The vader model uses a lexical approach, where lexical features are mapped to a sentiment score in a dictionary. This even includes colloquialisms such as "LOL" or ":/". The VADER sentiment dictionary was created by a collection of human raters and their average score for each word. Defining the sentiment of a sentence is done by adding the scores of the lexical units and normalizing the result.

VADERs 5 Heuristics:

1. Punctuation
   • Exclamantion marks change the intensity of a sentence
   • E.g. Exclamation marks amplify the sentiment of a sentence proportional to the number of exclamation marks
2. Capitalization
   • Capitalization changes the intensity of a word
   • E.g. The model increments or decrements the sentiment score of a word if it capitalized
3. Degree Modifiers
   • VADER maintians a booster dictionary, which can boost or dampen the score of a word
   • E.g. The "sort of" modifier would dampen the word "cool", "sort of cool"
4. Polarity shift
   • VADER essentially has a but checker
   • E.g. "I enjoy this but i would rather do that" shows a shift in polarity after the "but" keyword
5. examining the tri-gram before a sentiment-laden lexical feature to catch polarity negation
   • Negation is captured by multiplying the sentiment score of the sentiment-laden lexical feature by an empirically-determined value 0.74

NOTE: The VADER model is more effective on short texts

TLDR: The VADER model is a look up table, using 5 different heuristics to determine the sentiment of a sentence.