So, you want to get every Fibonaccith word? 1st, 2nd, 3rd, 5th, 8th, etc... You'd just write a tokenizer to split up the text into words, then iterate through them while keeping track of the current work number and the next fibonacci number, when they match, you spit out that word and calculate the next fibonacci number and move on. To calculate the next fibonacci number, you'd need to remember the previous one too, which is trivial.