When managing a dataset in which a matter corresponds to only one oracle doc, we use the 2nd approach: For each query, we get its corresponding doc as oracle doc and randomly decide on a specified variety of files from https://elainezynq069465.wikicorrespondence.com/user