load_data.py 文件源码

def prepare_split_vec_dataset(dataset, word_index, padding = True, prem_len = None, hypo_len = None):
    P = []
    H = []
    y = []
    for example in dataset:
        if example[2] == '-':
            continue

        P.append(load_word_indices(example[0], word_index))   
        H.append(load_word_indices(example[1], word_index))
        y.append(LABEL_LIST.index(example[2]))

    one_hot_y = np.zeros((len(y), len(LABEL_LIST)))
    one_hot_y[np.arange(len(y)), y] = 1
    if pad_sequences:
        P = pad_sequences(P, prem_len, padding='pre')
        H = pad_sequences(H, hypo_len, padding='post')
    return np.array(P), np.array(H), one_hot_y