corenlp_local.py

def get_g_wvi_bert_from_g_wvi_corenlp(wh_to_wp_index, g_wvi_corenlp):
    """
    Generate SQuAD style start and end index of wv in nlu. Index is for of after WordPiece tokenization.
    Assumption: where_str always presents in the nlu.
    """
    g_wvi = []
    for b, g_wvi_corenlp1 in enumerate(g_wvi_corenlp):
        wh_to_wp_index1 = wh_to_wp_index[b]
        g_wvi1 = []
        for i_wn, g_wvi_corenlp11 in enumerate(g_wvi_corenlp1):

            st_idx, ed_idx = g_wvi_corenlp11

            st_wp_idx = wh_to_wp_index1[st_idx]
            ed_wp_idx = wh_to_wp_index1[ed_idx]

            g_wvi11 = [st_wp_idx, ed_wp_idx]
            g_wvi1.append(g_wvi11)

        g_wvi.append(g_wvi1)

    return g_wvi

def get_g_wvi_corenlp(t):
    g_wvi_corenlp = []
    for t1 in t:
        g_wvi_corenlp.append( t1['wvi_corenlp'] )
    return g_wvi_corenlp