What's the best way to count keywords in JavaScript?

Question

What's the best and most efficient way to count keywords in JavaScript? Basically, I'd like to take a string and get the top N words or phrases that occur in the string, mainly for the use of suggesting tags. I'm looking more for conceptual hints or links to real-life examples than actual code, but I certainly wouldn't mind if you'd like to share code as well. If there are particular functions that would help, I'd also appreciate that.

Right now I think I'm at using the split() function to separate the string by spaces and then cleaning punctuation out with a regular expression. I'd also want it to be case-insensitive.

Answer 1

Cut, paste + execute demo:

var text = "Text to be examined to determine which n words are used the most";

// Find 'em!
var wordRegExp = /\w+(?:'\w{1,2})?/g;
var words = {};
var matches;
while ((matches = wordRegExp.exec(text)) != null)
{
    var word = matches[0].toLowerCase();
    if (typeof words[word] == "undefined")
    {
        words[word] = 1;
    }
    else
    {
        words[word]++;
    }
}

// Sort 'em!
var wordList = [];
for (var word in words)
{
    if (words.hasOwnProperty(word))
    {
        wordList.push([word, words[word]]);
    }
}
wordList.sort(function(a, b) { return b[1] - a[1]; });

// Come back any time, straaanger!
var n = 10;
var message = ["The top " + n + " words are:"];
for (var i = 0; i < n; i++)
{
    message.push(wordList[i][0] + " - " + wordList[i][1] + " occurance" +
                 (wordList[i][1] == 1 ? "" : "s"));
}
alert(message.join("\n"));

Reusable function:

function getTopNWords(text, n)
{
    var wordRegExp = /\w+(?:'\w{1,2})?/g;
    var words = {};
    var matches;
    while ((matches = wordRegExp.exec(text)) != null)
    {
        var word = matches[0].toLowerCase();
        if (typeof words[word] == "undefined")
        {
            words[word] = 1;
        }
        else
        {
            words[word]++;
        }
    }

    var wordList = [];
    for (var word in words)
    {
        if (words.hasOwnProperty(word))
        {
            wordList.push([word, words[word]]);
        }
    }
    wordList.sort(function(a, b) { return b[1] - a[1]; });

    var topWords = [];
    for (var i = 0; i < n; i++)
    {
        topWords.push(wordList[i][0]);
    }
    return topWords;
}

Answer 2

Once you have that array of words cleaned up, and let's say you call it wordArray:

var keywordRegistry = {};

for(var i = 0; i < wordArray.length; i++) {
   if(keywordRegistry.hasOwnProperty(wordArray[i]) == false) {
      keywordRegistry[wordArray[i]] = 0;
   }
   keywordRegistry[wordArray[i]] = keywordRegistry[wordArray[i]] + 1;
}

// now keywordRegistry will have, as properties, all of the 
// words in your word array with their respective counts 

// this will alert (choose something better than alert) all words and their counts
for(var keyword in keywordRegistry) {
  alert("The keyword '" + keyword + "' occurred " + keywordRegistry[keyword] + " times");
}

That should give you the basics of doing this part of the work.

Answer 3

Try to split you string on words and count the resulting words, then sort on the counts.

Answer 4

This builds upon a previous answer by insin by only having one loop:

function top_words(text, n) {
    // Split text on non word characters
    var words = text.toLowerCase().split(/\W+/)
    var positions = new Array()
    var word_counts = new Array()
    for (var i=0; i<words.length; i++) {
        var word = words[i]
        if (!word) {
            continue
        }

        if (typeof positions[word] == 'undefined') {
            positions[word] = word_counts.length
            word_counts.push([word, 1])
        } else {
            word_counts[positions[word]][1]++
        }
    }
    // Put most frequent words at the beginning.
    word_counts.sort(function (a, b) {return b[1] - a[1]})
    // Return the first n items
    return word_counts.slice(0, n)
}

// Let's see if it works.
var text = "Words in here are repeated. Are repeated, repeated!"
alert(top_words(text, 3))

The result of the example is: [['repeated',3], ['are',2], ['words', 1]]

Answer 5

I would do exactly what you have mentioned above to isolate each word. I would then probably add each word as the index of an array with the number of occurrences as the value.

For example:

var a = new Array;
a[word] = a[word]?a[word]+1:1;

Now you know how many unique words there are (a.length) and how many occurrences of each word existed (a[word]).