Abstract. We study the topmost weight matrix of neural network language models. We show that this matrix constitutes a valid word embedding.
確定！ 回上一頁
