N$-$fold cross$-$validation is a technique for evaluation that uses All data in the corpus for training$90\%$ of the data in the corpus for training $10\%$ of the data in the corpus for training $50\%$ of the data in the corpus for trainingQuestion $15(5$ points$)4$Listen When training a language model, if we use an overly narrow corpus, theprobabilities Reflect all possible wordingsDon't generalize Don't reflect the task Reflect intuition