datasets.py 文件源码

def _encode(self, s, V, context):
        """
        Arguments
        ----------
            s: Sentence as a list of strings
            V: Vocabulary as a np array of strings
            context: The maximum length of previous words to include
        """
        idxs = np.searchsorted(V, s)
        x = np.zeros((len(s)-1, context), dtype=np.int32)
        y = np.zeros((len(s)-1, 1), np.int32)
        for i in range(1, len(s)):
            x[i-1, :i] = idxs[:i][-context:] + 1  # 0 means missing value
            y[i-1] = idxs[i]

        return x, y