尝试在nltk_contrib.timex中使用地面方法时出现错误

Question

import datetime
from nltk_contrib import timex

now = datetime.date.today()
basedate = timex.Date(now.year, now.month, now.day)

print timex.ground(timex.tag("Hai i would like to go to mumbai 22nd of next month"), basedate)

print str(datetime.date.day)

当我尝试运行上面的代码时，我得到了以下错误

File "/usr/local/lib/python2.7/dist-packages/nltk_contrib/timex.py", line 250, in ground
    elif re.match(r'last ' + month, timex, re.IGNORECASE):
UnboundLocalError: local variable 'month' referenced before assignment

我应该做些什么来纠正这个错误？

Answer 1

timex模块有一个错误，即在ground函数中引用全局变量时不进行赋值。

要修复该错误，请添加以下代码，该代码应从第171行开始：

定义地面(tagged_text、base_date)：

# Find all identified timex and put them into a list
timex_regex = re.compile(r'<TIMEX2>.*?</TIMEX2>', re.DOTALL)
timex_found = timex_regex.findall(tagged_text)
timex_found = map(lambda timex:re.sub(r'</?TIMEX2.*?>', '', timex), \
            timex_found)

# Calculate the new date accordingly
for timex in timex_found:
    global month # <--- here is the global reference assignment

Answer 2

当timex被连续多次调用时，上面关于将月添加为全局变量的解决方案会导致其他问题，因为除非您再次导入，否则变量不会重置。在AWS Lambda的部署环境中，这会发生在我身上。

一个不是很漂亮但不会造成问题的解决方案是在ground函数中再次设置Month值：

def ground(tagged_text, base_date):

    # Find all identified timex and put them into a list
    timex_regex = re.compile(r'<TIMEX2>.*?</TIMEX2>', re.DOTALL)
    timex_found = timex_regex.findall(tagged_text)
    timex_found = map(lambda timex:re.sub(r'</?TIMEX2.*?>', '', timex), \
            timex_found)

    # Calculate the new date accordingly
    for timex in timex_found:
        month = "(january|february|march|april|may|june|july|august|september| \
        october|november|december)" # <--- reset month to the value it is set to upon import