Adding words to improve classification

I am familiar with basic classification (given a span of text, what label should it get) but I am wondering about scenarios where there is more than just a span of text to use. Let’s imagine a scenario where in addition to the text, there is also a date, location, and time. If one (or all) of them were appended on to the text, would this hurt or help the classification? I can see it going both ways. Has anyone tried doing something like this?

Another similar scenario would be for tweet classification. The tweet contains emojis/emoticons and urls, so every time there is an emoji/emoticon, a description of the face or emotion replaces the emoji/emoticon (:slight_smile: becomes “happy”, :frowning: becomes “sad”) and then the url (https://www…) gets turned into the word “url” or “link.” Does this information hurt or help the classifier?

I will probably try it out myself, but I’d like to hear from others for their perspective or experience.