Google Colab安装Parsivar Normalizer失败求助

问题描述 #

我在Google Colab中安装Parsivar Normalizer时遇到错误，该错误已出现约两周，此前安装无问题。

安装代码：

!pip install parsivar
from parsivar import Normalizer
my_normalizer = Normalizer()

报错调用栈：

ImportError                               Traceback (most recent call last)
<ipython-input-1-a4ae8568f137> in <cell line: 2>()
      1 get_ipython().system('pip install parsivar')
----> 2 from parsivar import Normalizer
      3 my_normalizer = Normalizer()

7 frames
/usr/local/lib/python3.10/dist-packages/parsivar/__init__.py in <module>
----> 1 from .chunker import FindChunks
      2 from .data_helper import DataHelper
      3 from .dependency import DependencyParser
      4 from .normalizer import Normalizer
      5 from .postagger import POSTagger

/usr/local/lib/python3.10/dist-packages/parsivar/chunker.py in <module>
----> 1 import nltk
      2 
      3 class FindChunks():
      4     def __init__(self):
      5         self.grammar = r"""

/usr/local/lib/python3.10/dist-packages/nltk/__init__.py in <module>
    132 from nltk.grammar import *
    133 from nltk.probability import *
--> 134 from nltk.text import *
    135 from nltk.tree import *
    136 from nltk.util import *

/usr/local/lib/python3.10/dist-packages/nltk/text.py in <module>
     24 from six import text_type
     25 
--> 26 from nltk.lm import MLE
     27 from nltk.lm.preprocessing import padded_everygram_pipeline
     28 from nltk.probability import FreqDist

/usr/local/lib/python3.10/dist-packages/nltk/lm/__init__.py in <module>
    220 """
    221 
--> 222 from nltk.lm.models import (
    223     MLE,
    224     Lidstone,

/usr/local/lib/python3.10/dist-packages/nltk/lm/models.py in <module>
     10 
     11 from nltk import compat
--> 12 from nltk.lm.api import LanguageModel, Smoothing
     13 from nltk.lm.smoothing import KneserNey, WittenBell
     14 

/usr/local/lib/python3.10/dist-packages/nltk/lm/api.py in <module>
     15 from six import add_metaclass
     16 
--> 17 from nltk.lm.counter import NgramCounter
     18 from nltk.lm.util import log_base2
     19 from nltk.lm.vocabulary import Vocabulary

/usr/local/lib/python3.10/dist-packages/nltk/lm/counter.py in <module>
     13 from __future__ import unicode_literals
     14 
--> 15 from collections import Sequence, defaultdict
     16 
     17 from six import string_types

ImportError: cannot import name 'Sequence' from 'collections' (/usr/lib/python3.10/collections/__init__.py)

错误原因 #

该错误源于Python版本兼容性问题：

• Google Colab当前默认使用Python 3.10，而Python 3.3+已将Sequence这类抽象基类从collections模块迁移至collections.abc模块；
• Parsivar依赖的旧版本NLTK仍在代码中从collections导入Sequence，导致在Python 3.10环境下触发ImportError。

解决方案 #

方案1：升级NLTK到兼容版本 #

NLTK 3.8及以上版本已修复此导入问题，直接升级即可：

!pip install --upgrade nltk
!pip install parsivar
from parsivar import Normalizer
my_normalizer = Normalizer()

方案2：手动修复NLTK代码（临时方案） #

如果升级后仍有问题，可手动修改NLTK的导入逻辑：

1. 运行以下命令定位NLTK文件路径：

import nltk
print(nltk.__file__)

2. 找到nltk/lm/counter.py文件，将第15行代码：

from collections import Sequence, defaultdict

修改为：

from collections.abc import Sequence
from collections import defaultdict

3. 保存后重新运行导入代码。

方案3：指定Parsivar兼容版本 #

若无需最新版Parsivar，可尝试安装旧版本搭配兼容的NLTK：

!pip install parsivar==0.1.9
!pip install nltk==3.7
from parsivar import Normalizer
my_normalizer = Normalizer()

内容的提问来源于stack exchange，提问作者HosseinSedghian