By Arvind Ravulavaru

Creating a Sentiment Analysis Application Using Node.js

By Arvind Ravulavaru

Sentiment analysis helps you in understanding what people think about a certain topic. The sentiment analysis app we are going to build will take in a keyword(s) and will fetch the matching tweets from Twitter. Then, it will run each tweet through an AFINN powered sentiment analysis Node module. This module will score the tweet text, and finally will display the relevant statistics.

Before we get started, you can take a look at the demo here. You can download the complete source code from here.

Technology Stack

This app is built on top of Nodejs. We will be using Express as our server side framework and jQuery for DOM manipulation on the client side. To make things easy for us, we will be using a Slush generator named slush-express (written by me) to scaffold a new Express app for us. We will be using a light weight disk based DB named diskDB (also written by me) to store the results.

We will be using the Twitter module to interact with Twitter and Sentiment module to perform the analysis.

So, let’s get started.

Setup the App

Create a new folder named sentimentAnalysisApp and open a new terminal/prompt here. First, we are going to install Gulp, Slush, and Express generator using the following command.

npm i -g gulp slush slush-express

Once the above modules are installed, we will scaffold a new Express app using this command:

slush express

Slush will ask you to pick a view engine and a style sheet engine. Answer as shown below.

[?] Select a View Engine: HTML
[?] Select a Stylesheet Engine: CSS

It will take slush a couple of minutes to scaffold the application and install the dependencies. The resulting scaffolded folder structure will look like this:

├── Gulpfile.js
├── app.js
├── bin
│   └── www
├── bower.json
├── package.json
├── public
│   └── stylesheets
│       └── style.css
├── routes
│   ├── index.js
│   └── users.js
└── views
    ├── error.html
    └── index.html

Here is a quick explanation of the different files and folders.

  • bin/www – The server initiation and the port is configured here.
  • app.js – The server configuration, the routes, and the view engine is configured here.
  • gulpFile.js – The task runner for our project.
  • /public – Consists of the static files dispatched to the UI.
  • /routes – Consists of the app routes.
  • /views – Consists of the app views.

You can run the app by executing the command gulp. This will start the Express server on port 3000. Navigate to http://localhost:3000 and you can see the sample home page.

Server Side Development

First, we will install the application level dependencies using the following command:

npm i twitter sentiment --save

Next, create a new folder named logic at the root of the project. Create two files named twitterSearch.js and sentimentAnalysis.js. These files consist of the logic for fetching tweets from Twitter and performing analysis, respectively. Create another folder named db, which will store the data.

Next, open routes/index.js in your favorite editor. We will add a new route, POST /search. The user entered search text will be submitted to this end point. Update routes/index.js as shown below.

'use strict';
var express = require('express');
var router = express.Router();
var twitterSearch = require('../logic/twitterSearch');

/* GET home page. */
router.get('/', function(req, res) {
});'/search', function(req, res) {
  twitterSearch(, function (data) {

module.exports = router;

The twitterSearch() function is expected to take in the search terms and fetch matching tweets from Twitter. These tweets will then be fed into the sentiment analysis module and the results will be returned as a callback. Simple, right?

Next, open logic/twitterSearch.js and add the following code.

var util = require('util'),
    twitter = require('twitter'),
    sentimentAnalysis = require('./sentimentAnalysis'),
    db = require('diskdb');

db = db.connect('db', ['sentiments']);
var config = {
  consumer_key: 'xxxxxxxxxxxxxxxxxxxxxxxxxxx',
  consumer_secret: 'xxxxxxxxxxxxxxxxxxxxxxxxxxx',
  access_token_key: 'xxxxxxxxxxxxxxxxxxxxxxxxxxx-xxxxxxxxxxxxxxxxxxxxxxxxxxx',
  access_token_secret: 'xxxxxxxxxxxxxxxxxxxxxxxxxxx'

module.exports = function(text, callback) {
  var twitterClient = new twitter(config);
  var response = [], dbData = []; // to store the tweets and sentiment, function(data) {
    for (var i = 0; i < data.statuses.length; i++) {
      var resp = {};

      resp.tweet = data.statuses[i];
      resp.sentiment = sentimentAnalysis(data.statuses[i].text);
        tweet: resp.tweet.text,
        score: resp.sentiment.score

For us to fetch the tweets, we need to first register a new app with Twitter. Then we will generate the required keys and tokens to be used while making a request.

Navigate to Twitter apps and click on Create New App. Fill the form on the next page. Once the app is created, we will generate the required tokens. Click on the API Keys tabs, and scroll down to the bottom of the page. There click on Create my access token. Once this is done, you will see a message to refresh the page, go ahead and do that. Now you can see an API key, API secret, Access token, and Access token secret and fill the config object like this:

var config = {
  consumer_key: 'API key',
  consumer_secret: 'API secret',
  access_token_key: 'Access token',
  access_token_secret: 'Access token secret'

If you face issues, please refer to this discussion.

Next, open logic/sentimentAnalysis.js and add the following code.

var sentiment = require('sentiment');

module.exports = function(text) {
  return sentiment(text);

The logic is very simple. we take in the tweet text and return the sentiment object. This wraps up our server. Now, let us build the client.

Client Side Development

First create a new folder named scripts inside the public folder. Inside scripts, create a new file named app.js and open it up in your favorite editor. app.js is responsible for submitting the form via Ajax to the /search endpoint and displaying the results on the page.

We are going to use John Resig’s JavaScript Micro-Templating library for building the markup from server data (a small scale MV* if you may). I have modified the templating library, so that we can use {{ }} instead of <% %> syntax. This way we can use ejs as our server side templating to render HTML extension pages. The completed app.js is shown below.

$(document).ready(function() {
  // handle the form submit
  $('#searchText').on('keypress', function(e) {
    if (e.which == 13 || e.keyCode == 13) {
      if ($(this).val().trim().length > 0) {
        // initiate an Ajax call to send the data

  function fireAJAX(text) {
      type: 'POST',
      url: '/search',
      data: {
        search: text
      beforeSend: function(xhr) {
      success: parseData,
      error: oops

  function parseData(data) {
    var html = '';
    for (var i = 0; i < data.length; i++) {
      var s = data[i].sentiment,
          t = data[i].tweet;

      var _o = {
        imgSrc: t.user.profile_image_url,
        tweetLink: '' + t.user.screen_name + '/status/' + t.id_str,
        tweet: t.text,
        score: s.score ? s.score : '--',
        comparative: s.comparative ? s.comparative : '--',
        favorited: t.favorite_count ? t.favorite_count : 0,
        retweet: t.retweet_count ? t.retweet_count : 0,
        wordsMatched: s.words && s.words.length ? s.words : '--',
        positiveWords: s.positive && s.positive.length ? s.positive : '--',
        negativeWords: s.negative && s.negative.length ? s.negative : '--'

      html += tmpl('tweet_tmpl', _o);

  function oops(data) {

  function disableState() {
    $('#searchText').prop('disabled', false);

  function enableState() {
    $('#searchText').prop('disabled', true);

// Simple JavaScript Templating
// John Resig - - MIT Licensed
(function() {
  var cache = {};

  this.tmpl = function tmpl(str, data) {
    // Figure out if we're getting a template, or if we need to
    // load the template - and be sure to cache the result.
    var fn = !/\W/.test(str) ?
      cache[str] = cache[str] ||
      tmpl(document.getElementById(str).innerHTML) :

    // Generate a reusable function that will serve as a template
    // generator (and which will be cached).
    new Function("obj",
      "var p=[],print=function(){p.push.apply(p,arguments);};" +

      // Introduce the data as local variables using with(){}
      "with(obj){p.push('" +

      // Convert the template into pure JavaScript
      .replace(/[\r\t\n]/g, " ")
      .split("{{").join("\t") // modified
      .replace(/((^|\}\})[^\t]*)'/g, "$1\r") // modified
      .replace(/\t=(.*?)}}/g, "',$1,'") // modified
      .split("}}").join("p.push('") // modified
      .split("\r").join("\\'") + "');}return p.join('');");

    // Provide some basic currying to the user
    return data ? fn(data) : fn;

Next, open views/index.html and add the following code.

<!DOCTYPE html>

  <title>Sentiment Analysis App</title>
  <link rel='stylesheet' href='/stylesheets/style.css' />

  <h1>Sentiment Analysis App</h1>
  <input type="text" id="searchText" placeholder="Enter the text you would like to see the analysis for and hit return">
  <div class="results">
    <lable class="loading">Loading.. Please wait</lable>
    <lable class="error">Oops.. Something went wrong</lable>
  <section class="tweet-results">

  <script src="//"></script>
  <script type="text/javascript" src="scripts/app.js"></script>
  <script type="text/html" id="tweet_tmpl">
      <div class="left">
        <img src="{{=imgSrc}}">
        <a href="{{=tweetLink}}" target="_blank">Link</a></p>
      <div class="right">
            <td>Words Matched</td>
            <td>Positive Words</td>
            <td>Negative Words</td>

Here, we are referring jQuery and app.js. We have also created a template (tweet_tmpl), that will be used to show the results. Finally, open stylesheets/style.css and add the following classes.

body {
  padding: 50px;
  font: 14px"Lucida Grande", Helvetica, Arial, sans-serif;
  background: #eee;
a {
  color: #00B7FF;
input {
  width: 98%;
  padding: 9px;
  font-size: 17px;
.results {
  display: none;
.error {
  color: red;
  display: none;
.tweet-results {
  width: 100%;
  overflow: hidden;
  padding-right: 18px;
.left {
  float: left;
  width: 39%;
.right {
  float: right;
  width: 55%;
  border-left: 1px dashed;
  padding-left: 21px;
article {
  background: #fff;
  display: block;
  padding: 18px;
  border: 1px solid #eee;
  margin-top: 21px;
  margin-bottom: 21px;
  overflow: hidden;
  box-shadow: 6px 4px 9px 1px rgba(119, 119, 119, 0.75);
  -moz-box-shadow: 6px 4px 9px 1px rgba(119, 119, 119, 0.75);
  -webkit-box-shadow: 6px 4px 9px 1px rgba(119, 119, 119, 0.75);
article img {
  width: 64px;
  float: left;
  margin:0 5px 0 0;
.right table {
  width: 100%;
.right table, .right table td {
  border: 1px solid;
.right table td {
  width: 50%;

That is it, we are done with our development. Let us run the app and test it. Back in the terminal, run the gulp command to start the server. Navigate to http://localhost:3000/ and you should see the search bar. Enter “This is Awesome” and hit return and you should see something like this:

Positive Results

Here, Score is the sum of scores for each word present in the tweet matching the AFINN source. Comparative is equal to score/total words. Words Matched show how many words from the tweet matched the AFINN words while processing. Positive Words are the matched positive words and Negative Words are the matched negative words. This data should give you enough information to make a decision and understand the sentiment.

Clear the search bar and enter sad broken and hit return. Your results should look something like this:

Negative Results

Simple and easy right? Now you can search a few words and see how the results turn out.


Training Your Data

You must have noticed by now that not all tweets return results. This is because, if none of the words in the tweet match the AFINN words, the score would be 0. For example:

Before Training

If you would like to fix this, you can train the sentiment module. Create a new file inside the logic folder named training.js and add the following code.

module.exports = {
  directives: 4,
  angular: 5,
  code: 3,
  scope: 3

Here, we are training the module to use the above scores for the mentioned words. Next, update logic/sentimentAnalysis.js like this:

var sentiment = require('sentiment');
var trainedData = require('./training.js');
module.exports = function(text) {
  return sentiment(text, trainedData);

After training, the results will look like this:

After Training

This is very a powerful technique, so make sure you assign proper values to the keywords, else you may see results which may not make sense. Refer to AFINN for further information.

Creating a RESTful Service

You can build a real time dashboard that will show tweets and results. You can run an async job, that will hit the Twitter API from time to time, fetch the data, and save it using diskDB. Then, you can expose this file as a RESTful end point. Add the following code to routes/index.js.

router.get('/data', function(req, res) {
    .connect('db', ['sentiments'])

Now, when you access http://localhost:3000/data you can see the complete data saved so far. A live example is available here.

Integrate Social Media

Similar to Twitter, you can integrate Facebook, Google+, and other social media data. All you need to do is pass the text you would like to analyze to logic/sentimentAnalysis.js, and you can see the score for it.


I hope you got a basic idea on how to perform sentiment analysis using this app. Thanks for reading! Do comment, please.

Login or Create Account to Comment
Login Create Account
Get the most important and interesting stories in tech. Straight to your inbox, daily.