WordRank

[Nathan Hughes]

Hopefully a quick easy query

Where 2 words have identical weight, how is rank decided? I can see any pattern.

Many Thanks

[Dan Bernier]

I had to look through the source to figure it out, but the answer is, it's arbitrary: it's decided by java internals.

If you look at https://github.com/danbernier/WordCram/blob/master/src/wordcram/WordCram.java#L786, you can see where it uses the WordSorterAndScaler to sort and scale the words ("scale," in this case, meaning to normalize the weights to 0..1).

https://github.com/danbernier/WordCram/blob/master/src/wordcram/WordSorterAndScaler.java#L13 shows that it sorts them via Array.sort(), relying on Word's default sorting criteria.

Finally, https://github.com/danbernier/WordCram/blob/master/src/wordcram/Word.java#L305 shows that it's strictly by weight - so words of equal weight are considered equal. This means the specifics are handled internally.

Just curious, what's your situation? Are these equal-weight words "heavy," or "light?"

[Nathan Hughes]

Hi

Yeah that was the point I got to as well.

What I'm trying to do is colour the top 5 words specific colours. I'm using this

class First5 implements WordColorer {

  int callCount = 0;

  public color colorFor(Word word) {

    callCount++;

    colorMode(RGB, 255, 255, 255);

    if (callCount == 1) {

      return color(2,27,65);

    } 

    else if (callCount == 2) {

     return color(0,53,173);

    } 

    else if (callCount == 3) {

     return color(0,163,224);

    } 

    else if (callCount == 4) {

     return color(209,216,227);

    } 

    else if (callCount == 5) {

     return color(204,51,51);

    } 

    else {

      return color(48,88,175);

    }

  }

}

Now while its possible to adapt this to colour words with the same weight the same way this could lead to dull wordclouds. So I just wanted to match the rank assigned by the array to my top 5 list so I could keep the colour list for the first 5 words, it doesn't seem to be byweight then by alphabetical or even byweight then by length, or even by the order the words are in the data file, it almost looks random.