From 25d53e77c057835c08122590cd8a1c90bb69c831 Mon Sep 17 00:00:00 2001 From: wl-zhao Date: Mon, 11 Mar 2024 02:14:05 +0000 Subject: [PATCH] improve split sentence --- melo/split_utils.py | 69 +++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 67 insertions(+), 2 deletions(-) diff --git a/melo/split_utils.py b/melo/split_utils.py index 6379be3..7f2bcf9 100644 --- a/melo/split_utils.py +++ b/melo/split_utils.py @@ -4,7 +4,7 @@ import glob import numpy as np import soundfile as sf import torchaudio -from txtsplit import txtsplit +import re def split_sentence(text, min_len=10, language_str='EN'): if language_str in ['EN', 'FR', 'ES', 'SP']: @@ -13,13 +13,15 @@ def split_sentence(text, min_len=10, language_str='EN'): sentences = split_sentences_zh(text, min_len=min_len) return sentences + def split_sentences_latin(text, min_len=10): text = re.sub('[。!?;]', '.', text) text = re.sub('[,]', ',', text) text = re.sub('[“”]', '"', text) text = re.sub('[‘’]', "'", text) text = re.sub(r"[\<\>\(\)\[\]\"\«\»]+", "", text) - return [item.strip() for item in txtsplit(text, 512, 512) if item.strip()] + return [item.strip() for item in txtsplit(text, 256, 512) if item.strip()] + def split_sentences_zh(text, min_len=10): text = re.sub('[。!?;]', '.', text) @@ -45,6 +47,7 @@ def split_sentences_zh(text, min_len=10): new_sent = [] return merge_short_sentences_zh(new_sentences) + def merge_short_sentences_en(sens): """Avoid short sentences by merging them with the following sentence. @@ -70,6 +73,7 @@ def merge_short_sentences_en(sens): pass return sens_out + def merge_short_sentences_zh(sens): # return sens """Avoid short sentences by merging them with the following sentence. @@ -97,6 +101,67 @@ def merge_short_sentences_zh(sens): return sens_out + +def txtsplit(text, desired_length=100, max_length=200): + """Split text it into chunks of a desired length trying to keep sentences intact.""" + text = re.sub(r'\n\n+', '\n', text) + text = re.sub(r'\s+', ' ', text) + text = re.sub(r'[""]', '"', text) + text = re.sub(r'([,.?!])', r'\1 ', text) + text = re.sub(r'\s+', ' ', text) + + rv = [] + in_quote = False + current = "" + split_pos = [] + pos = -1 + end_pos = len(text) - 1 + def seek(delta): + nonlocal pos, in_quote, current + is_neg = delta < 0 + for _ in range(abs(delta)): + if is_neg: + pos -= 1 + current = current[:-1] + else: + pos += 1 + current += text[pos] + if text[pos] == '"': + in_quote = not in_quote + return text[pos] + def peek(delta): + p = pos + delta + return text[p] if p < end_pos and p >= 0 else "" + def commit(): + nonlocal rv, current, split_pos + rv.append(current) + current = "" + split_pos = [] + while pos < end_pos: + c = seek(1) + if len(current) >= max_length: + if len(split_pos) > 0 and len(current) > (desired_length / 2): + d = pos - split_pos[-1] + seek(-d) + else: + while c not in '!?.\n ' and pos > 0 and len(current) > desired_length: + c = seek(-1) + commit() + elif not in_quote and (c in '!?\n' or (c in '.,' and peek(1) in '\n ')): + while pos < len(text) - 1 and len(current) < max_length and peek(1) in '!?.': + c = seek(1) + split_pos.append(pos) + if len(current) >= desired_length: + commit() + elif in_quote and peek(1) == '"' and peek(2) in '\n ': + seek(2) + split_pos.append(pos) + rv.append(current) + rv = [s.strip() for s in rv] + rv = [s for s in rv if len(s) > 0 and not re.match(r'^[\s\.,;:!?]*$', s)] + return rv + + if __name__ == '__main__': zh_text = "好的,我来给你讲一个故事吧。从前有一个小姑娘,她叫做小红。小红非常喜欢在森林里玩耍,她经常会和她的小伙伴们一起去探险。有一天,小红和她的小伙伴们走到了森林深处,突然遇到了一只凶猛的野兽。小红的小伙伴们都吓得不敢动弹,但是小红并没有被吓倒,她勇敢地走向野兽,用她的智慧和勇气成功地制服了野兽,保护了她的小伙伴们。从那以后,小红变得更加勇敢和自信,成为了她小伙伴们心中的英雄。" en_text = "I didn’t know what to do. I said please kill her because it would be better than being kidnapped,” Ben, whose surname CNN is not using for security concerns, said on Wednesday. “It’s a nightmare. I said ‘please kill her, don’t take her there.’"