From 6427c834dfd97b1f05c6659cdc7ccf010bf82fe1 Mon Sep 17 00:00:00 2001
From: 嘉渊 <wangjiaming.wjm@alibaba-inc.com>
Date: 星期一, 24 四月 2023 19:50:07 +0800
Subject: [PATCH] update
---
funasr/datasets/large_datasets/utils/tokenize.py | 43 ++++++++++++++++++++++++++++++++-----------
1 files changed, 32 insertions(+), 11 deletions(-)
diff --git a/funasr/datasets/large_datasets/utils/tokenize.py b/funasr/datasets/large_datasets/utils/tokenize.py
index 0c01885..d8ceff2 100644
--- a/funasr/datasets/large_datasets/utils/tokenize.py
+++ b/funasr/datasets/large_datasets/utils/tokenize.py
@@ -18,35 +18,56 @@
def seg_tokenize(txt, seg_dict):
out_txt = ""
- pattern = re.compile(r"([\u4E00-\u9FA5A-Za-z0-9])")
for word in txt:
- if pattern.match(word):
- if word in seg_dict:
- out_txt += seg_dict[word] + " "
- else:
- out_txt += "<unk>" + " "
+ if word in seg_dict:
+ out_txt += seg_dict[word] + " "
else:
- continue
+ out_txt += "<unk>" + " "
return out_txt.strip().split()
def tokenize(data,
vocab=None,
- seg_dict=None):
+ seg_dict=None,
+ punc_dict=None,
+ bpe_tokenizer=None):
assert "text" in data
assert isinstance(vocab, dict)
text = data["text"]
token = []
+ vad = -2
+
+ if bpe_tokenizer is not None:
+ text = bpe_tokenizer.text2tokens("".join(text))
if seg_dict is not None:
assert isinstance(seg_dict, dict)
txt = forward_segment("".join(text).lower(), seg_dict)
text = seg_tokenize(txt, seg_dict)
-
- for x in text:
- if x in vocab:
+
+ length = len(text)
+ for i in range(length):
+ x = text[i]
+ if i == length-1 and "punc" in data and text[i].startswith("vad:"):
+ vad = x[-1][4:]
+ if len(vad) == 0:
+ vad = -1
+ else:
+ vad = int(vad)
+ elif x in vocab:
token.append(vocab[x])
else:
token.append(vocab['<unk>'])
+ if "punc" in data and punc_dict is not None:
+ punc_token = []
+ for punc in data["punc"]:
+ if punc in punc_dict:
+ punc_token.append(punc_dict[punc])
+ else:
+ punc_token.append(punc_dict["_"])
+ data["punc"] = np.array(punc_token)
+
data["text"] = np.array(token)
+ if vad is not -2:
+ data["vad_indexes"]=np.array([vad], dtype=np.int64)
return data
--
Gitblit v1.9.1