vader-sentiment-analysis

Installation
SKILL.md

VADER Sentiment Analysis

Overview

VADER (Valence Aware Dictionary and sEntiment Reasoner) is a lexicon-and-rule-based sentiment tool optimized for social media and informal text. It produces four scores per text: positive, negative, neutral (proportions summing to 1.0), and a compound score normalized to [-1, +1]. The core principle: VADER reads sentiment signals that traditional NLP preprocessing destroys -- capitalization for emphasis, punctuation for intensity, emoticons for affect -- so preprocessing must preserve these signals, not strip them.

Compound score normalization formula:

S_compound = sum(V_i) / sqrt(sum(V_i)^2 + alpha)

Where V_i are the valence scores of each token after rule-based adjustments (negation, degree modifiers, capitalization boost, punctuation amplification), and alpha = 15 is the normalization constant. This bounds the score to [-1, +1] with diminishing sensitivity at extremes.

When to Use

Installs
1
GitHub Stars
7
First Seen
Jul 5, 2026
vader-sentiment-analysis — aaddrick/written-voice-replication