LCA_LLM_application/Retrieval_new/utils.py

201 lines
6.0 KiB
Python
Raw Permalink Normal View History

2025-03-19 21:02:49 +08:00
import nltk
from nltk.tokenize import word_tokenize
from nltk import pos_tag
import jieba.posseg as pseg
2025-03-21 16:09:46 +08:00
# # 下载相关数据
# nltk.download('punkt')
# nltk.download('averaged_perceptron_tagger')
2025-03-19 21:02:49 +08:00
from nltk.stem import WordNetLemmatizer
import string
import re
2025-03-21 16:09:46 +08:00
from langchain.prompts import ChatPromptTemplate
from langchain.schema import SystemMessage, HumanMessage
from langchain_openai import ChatOpenAI
2025-03-31 10:07:03 +08:00
import logging
from typing import Optional
import time
# 配置日志
logging.basicConfig(
level=logging.INFO,
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
)
logger = logging.getLogger('translation_service')
2025-03-19 21:02:49 +08:00
def preprocess_eng(text):
'''
英文文本预处理小写化去除标点待定去除特殊符号只保留单词
拼写是否正确因为是从ecoinvent导入的没有拼写错误
词干提取(stemming)和词形还原(lemmatization)可以处理一下有的提取不准确不做此操作
'''
# 去除标点
text = text.translate(str.maketrans('', '', string.punctuation))
# 去除数字
text = re.sub(r'\d+', ' ', text)
# 去除多余字符
text = re.sub(r'[^A-Za-z0-9\s]', '', text)
# 去除多余空格
text = re.sub(r'\s+', ' ', text)
return text
def preprocess_zh(text):
'''
中文文本预处理:只保留中文内容去除英文数字和标点
'''
text = str(text)
# 去除英文
text = re.sub(r'[a-zA-Z]',' ',text)
text = re.sub(r'\d', ' ', text)
# 去除中文标点符号
text = re.sub(r'[,。!?、;:“”()《》【】-]', ' ', text)
# 去除英文标点符号
text = re.sub(r'[.,!?;:"\'\(\)\[\]{}]', ' ', text)
# 去除空格
text = re.sub(r'\s+','',text)
return text
# 英文名词处理
def get_noun_en(text):
# 分词
words = word_tokenize(text)
# 词性标注
tagged = pos_tag(words)
# 提取名词
nouns = [word for word, tag in tagged if tag.startswith('NN')]
noun = ' '.join(nouns)
return noun
# 中文名词提取
def get_noun_zh(text):
x = str(text)
if x=='nan':
return ''
words = pseg.cut(text)
nouns = [word for word, flag in words if flag.startswith('n')]
noun = ' '.join(nouns)
return noun
2025-03-31 10:07:03 +08:00
def all_chinese(text):
2025-03-19 21:02:49 +08:00
"""
判断一个文本是否不包含中文字符
参数:
text (str): 需要检查的文本
返回:
bool: 如果文本中没有中文字符返回True否则返回False
"""
for char in text:
if '\u4e00' <= char <= '\u9fff' or \
'\u3400' <= char <= '\u4dbf' or \
'\u2f00' <= char <= '\u2fdf' or \
'\u3100' <= char <= '\u312f' or \
'\u31a0' <= char <= '\u31bf':
2025-03-31 10:07:03 +08:00
flag = 1
else:
2025-03-19 21:02:49 +08:00
return False
2025-03-21 16:09:46 +08:00
return True
2025-03-31 10:07:03 +08:00
def extract_list(text: str) -> Optional[str]:
"""从文本中提取方括号内的内容"""
if not isinstance(text, str):
return None
2025-03-21 16:09:46 +08:00
try:
2025-03-31 10:07:03 +08:00
pattern = r'\[(.*?)\]'
matches = re.findall(pattern, text)
if not matches:
return None
2025-03-21 16:09:46 +08:00
return matches[-1]
except Exception as e:
2025-03-31 10:07:03 +08:00
logger.error(f"字符串处理异常: {e}")
2025-03-21 16:09:46 +08:00
return None
2025-03-31 10:07:03 +08:00
def translate(query: str) -> Optional[str]:
"""
将查询中的英文翻译为中文
如果提取列表为空最多重试三次
"""
if not query or not isinstance(query, str):
return None
2025-03-21 16:09:46 +08:00
sys_template = '''
你是一个专注于化工环境学科领域的翻译专家
用户将提供一个生命周期评价领域数据库的查询查询可能包含中英文字符你的任务是
1. 将查询中的所有英文表述转化为对应的中文表述
2. 确保转化后的查询中不含任何非中文语言
2025-03-31 10:07:03 +08:00
3. 将完整的中文查询以"[]"格式返回
4. 不返回除"[]"格式外的任何其他内容
2025-03-21 16:09:46 +08:00
请严格按照上述要求执行
'''
human_template = "查询内容为:{context}"
chat_prompt = ChatPromptTemplate.from_messages([
("system", sys_template),
("human", human_template)
])
messages = chat_prompt.format_messages(context=query)
2025-03-31 10:07:03 +08:00
2025-03-21 16:09:46 +08:00
llm = ChatOpenAI(
2025-03-31 10:07:03 +08:00
model="deepseek-chat",
2025-03-21 16:09:46 +08:00
base_url="https://api.deepseek.com",
api_key="sk-3e42e538bc39411ab80761106d83dda9",
temperature=0,
)
2025-03-31 10:07:03 +08:00
# 最多尝试三次
max_attempts = 3
for attempt in range(1, max_attempts + 1):
logger.info(f"翻译尝试 {attempt}/{max_attempts}{query[:50]}{'...' if len(query) > 50 else ''}")
try:
# 调用API获取翻译结果
response = llm.invoke(messages)
content = response.content
# 尝试提取结果
result = extract_list(content)
# 如果成功提取到结果,直接返回
if result is not None:
logger.info(f"成功提取翻译结果 (尝试 {attempt}/{max_attempts})")
return result
# 提取失败,记录信息
logger.warning(f"未能提取翻译结果 (尝试 {attempt}/{max_attempts}): {content[:100]}")
# 如果已经是最后一次尝试则返回None
if attempt == max_attempts:
logger.error("所有尝试均失败,无法获取有效翻译结果")
return None
# 短暂等待后继续下一次尝试
time.sleep(1)
except Exception as e:
logger.error(f"翻译过程中发生异常 (尝试 {attempt}/{max_attempts}): {e}")
if attempt == max_attempts:
return None
return None
# 使用示例
if __name__ == "__main__":
query = "HCOOH"
result = translate(query)
if result:
print(f"翻译结果: {result}")
else:
print("翻译失败")
2025-03-21 16:09:46 +08:00