ABSTRACT
I want to model and store the data so that it isamenable to future retrieval in all sorts of ways. For the term-frequency task, if it’s done several times, we might not always want to read and parse the file in its raw form every time we want to compute the term frequency. Also, we might want to mine for more facts about the books, not just term frequencies. So, instead of consuming the data in its raw form, this style encourages using alternative representations of the input data that make it easier to mine, now and in the future. In one approach to such a goal, the types of data (domains) that need to be stored are identified, and pieces of concrete data are associated with those domains, forming tables. With an unambiguous entity-relationship model in place, we can then fill in the tables with data, so to retrieve portions of it using declarative queries.
