GitHub - hssfloat/sentiment: AFINN-based sentiment analysis for Node.js.

AFINN-based sentiment analysis for Node.js

Sentiment is a Node.js module that uses the AFINN-111 wordlist to perform sentiment analysis on arbitrary blocks of input text. Sentiment provides serveral things:

A fully async interface for performing sentiment analysis
A build process that makes updating sentiment to future versions of the AFINN word list trivial
Performance (see benchmarks below)

Installation

Usage

var sentiment = require('sentiment');

sentiment('Cats are stupid.', function (err, result) {
    console.dir(result);    // Score: -2, Comparative: -0.666
});

sentiment('Cats are totally amazing!', function (err, result) {
    console.dir(result);    // Score: 4, Comparative: 1
});

Adding / overwriting words

You can append and/or overwrite values from AFINN by simply injecting key/value pairs into a sentiment method call:

var sentiment = require('sentiment');

sentiment('Cats are totally amazing!', {
    'cats': 5,
    'amazing': 2  
}, function (err, result) {
    console.dir(result);    // Score: 7, Comparative: 1.75
});

Testing

Benchmarks

The primary motivation for designing sentiment was performance. As such, sentiment includes a benchmark script within the test directory that compares it against the Sentimental module which provides a nearly equivalent interface and approach. Based on these benchmarks, running on an older MacBook Air with Node 0.8.9, sentiment is about 33 times faster than the alternative implementation:

sentiment (v0.1.0)
1000 operations  |  33ms
2000 operations  |  75ms
4000 operations  |  138ms
8000 operations  |  252ms
16000 operations |  506ms
32000 operations |  1120ms
64000 operations |  2268ms

--------------------------
--------------------------

Sentimental (v0.0.2)
1000 operations  |  1189ms
2000 operations  |  2392ms
4000 operations  |  4760ms
8000 operations  |  9469ms
16000 operations |  19513ms
32000 operations |  41018ms
64000 operations |  86931ms