Calculating Keyword Prominence isn’t hard
At TigerStep, we believe that although we invented Keyword Prominence ™, everyone should be able to benefit from it. So here’s how you calculate it. First, work out keyword densities, in the traditional oldskool way that Keyword Prominence replaced. Associate all keyphrases with lexically identical phrases, which are words with identical stems, plurals etc. Sum the word densities for those associated groups. Multiply that score by the number of words in the original keyphrase. To make the resulting list of words easy to order, modify the result slightly by multiplying by the word length (we use a tiny mod of 0.001). Next, modify the score by the word’s position within the overall text. To do this, find the average position in the text of that key (a 100-word text with the key at 1 and 100 would be (100 + 1)/2, for example). Multiply the existing score by that weight, adjusted to mean that a perfectly distributed keyword isn’t really changed, but keywords clustered towards the start of the text have higher prominence and those clustered towards the end have lower keyword prominence. We effectively add 0.5 to the weight to normalise it in this way. Next, adjust the score to reflect the number of decorations around that keyword. Decorations are effects that make the keyword more noticeable:- bold, italic, headings, links, and block quotes for example. These enhancements draw the eye, so making the keyword literally more ‘prominent’. We use a simple and pragmatic way to factor in these enhancements by counting the decorations, multiplying by the number of words in the keyphrase, dividing by 0.5, and then adding the result to the existing score. Next, check to see if the exact keyphrase is in the text’s title (only applicable to blog posts). if it is, it’s worth 2 extra occurrences of the keyphrase for each time it appears as a separate phrase. If it is present in a lexically identical way, it’s worth the same as an extra occurrence in the text. This is necessary because search engines place heavy weight on the title of a post – it’s prominent on their search results, as well as on the page itself, and people often search by exact phrase nowadays. All that’s left is to check any images in the text for occurrences of key phrases, either as the filename of the image, or in any of the image’s other properties, such as the alt text. The search engines undoubtedly use sophisticated image analysis strategies to determine what the topic of an image is:- we can only do what the engines used to do, and use filenames and image meta properties as a proxy. This result is added to the score after being multiplied by 0.5 times the key’s word count, to reflect the fact that it’s not as significant as actual text.