A data scientist computes TF-IDF for the word "neural" in a corpus of 1,000 documents. It appears in 500 documents and occurs 10 times in a 200-word document. A colleague says "TF-IDF will be high because the word appears 10 times." What is wrong with this reasoning?