Browse Source

增加语义切分模型 (#248)

royd 2 years ago
parent
commit
23a6b26f3e
1 changed files with 21 additions and 8 deletions
  1. 21 8
      textsplitter/chinese_text_splitter.py

+ 21 - 8
textsplitter/chinese_text_splitter.py

@@ -1,25 +1,38 @@
 from langchain.text_splitter import CharacterTextSplitter
 import re
 from typing import List
+from modelscope.pipelines import pipeline
 
 
+p = pipeline(
+    task="document-segmentation",
+    model='damo/nlp_bert_document-segmentation_chinese-base',
+    device="cpu")
+
 class ChineseTextSplitter(CharacterTextSplitter):
     def __init__(self, pdf: bool = False, **kwargs):
         super().__init__(**kwargs)
         self.pdf = pdf
 
-    def split_text(self, text: str) -> List[str]:
+    def split_text(self, text: str, use_document_segmentation: bool=False) -> List[str]:
+        # use_document_segmentation参数指定是否用语义切分文档,此处采取的文档语义分割模型为达摩院开源的nlp_bert_document-segmentation_chinese-base,论文见https://arxiv.org/abs/2107.09278
+        # 如果使用模型进行文档语义切分,那么需要安装modelscope[nlp]:pip install "modelscope[nlp]" -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html
+        # 考虑到使用了三个模型,可能对于低配置gpu不太友好,因此这里将模型load进cpu计算,有需要的话可以替换device为自己的显卡id
         if self.pdf:
             text = re.sub(r"\n{3,}", "\n", text)
             text = re.sub('\s', ' ', text)
             text = text.replace("\n\n", "")
-        sent_sep_pattern = re.compile('([﹒﹔﹖﹗.。!?]["’”」』]{0,2}|(?=["‘“「『]{1,2}|$))')  # del :;
-        sent_list = []
-        for ele in sent_sep_pattern.split(text):
-            if sent_sep_pattern.match(ele) and sent_list:
-                sent_list[-1] += ele
-            elif ele:
-                sent_list.append(ele)
+        if use_document_segmentation:
+            result = p(documents=text)
+            sent_list = [i for i in result["text"].split("\n\t") if i]
+        else:
+            sent_sep_pattern = re.compile('([﹒﹔﹖﹗.。!?]["’”」』]{0,2}|(?=["‘“「『]{1,2}|$))')  # del :;
+            sent_list = []
+            for ele in sent_sep_pattern.split(text):
+                if sent_sep_pattern.match(ele) and sent_list:
+                    sent_list[-1] += ele
+                elif ele:
+                    sent_list.append(ele)
         return sent_list