|
I once heard that TF/IDF was developed from the field of information retrieval. It is not very appropriate for text classification purposes. I am not sure how to understand this statement correctly. |
|
I don't agree with that statement. It is common to apply TF/IDF before doing text classification. |
|
I'm not sure I understand the question here...TF/IDF is a way to quantify how significant a particular word is to a particular document in a given document set. Perhaps, that's why whoever it was told you that it's not good in text classification. However, you could use TF/IDF as a feature in text classification. For example, in spam filtering, you could use TF/IDF of each word in an email as a feature, or you could use some function and/or aggregation of the TF/IDF of each word in an email... |
