Half-OT: Note that words such as "like" and "um" are typically not referred to as stop words in (computational) linguistics. They are called filled pauses, or sometimes floor holders, depending on their function. Liz Shriberg's thesis was - as far as I know - the first extensive treatment of speech disfluencies in the field of computational linguistics.
The term "stop words" is, however, used in information retrieval. Here, it refers to words that appear very frequently in (almost) all documents in a corpus; so frequently in fact that they are taken not to carry any/much information content at all. Examples would be "the", "it", "and" etc. For many tasks in IR, stop words are removed from the document representation because they mostly introduce noise.
The term "stop words" is, however, used in information retrieval. Here, it refers to words that appear very frequently in (almost) all documents in a corpus; so frequently in fact that they are taken not to carry any/much information content at all. Examples would be "the", "it", "and" etc. For many tasks in IR, stop words are removed from the document representation because they mostly introduce noise.