The retrieved documents, from Step 4, function as
input to this step. Either full text documents or the
snippets (document summaries from the search
engines) can be used. Further, the full text
documents or the snippets can either be processed
by creating document feature vectors (DFV) or
keeping the texts in their raw form (e.g., the text
without HTML tags). However, the question is,
does the FV quality improve by using full text
documents compared to using only snippets?