计算文件中的字符串，一些单个单词，一些完整的句子

Question

我想计算文件中某些单词和名称的出现次数。下面的代码错误地将 fish and chips 计为 fish 的一种情况和 chips 的一种情况，而不是 fish and chips.

的一种计数

ngh.txt = 'test file with words fish, steak fish chips fish and chips'

import re
from collections import Counter
wanted = '''
"fish and chips"
fish
chips
steak
'''
cnt = Counter()
words = re.findall('\w+', open('ngh.txt').read().lower())
for word in words:
    if word in wanted:
        cnt[word] += 1
print cnt

输出：

Counter({'fish': 3, 'chips': 2, 'and': 1, 'steak': 1})

我想要的是：

Counter({'fish': 2, 'fish and chips': 1, 'chips': 1, 'steak': 1})

（理想情况下，我可以获得这样的输出：

fish: 2
fish and chips: 1
chips: 1
steak: 1

)

Answer 1

因此此解决方案适用于您的测试数据（并在测试数据中添加了一些术语，只是为了彻底），尽管它可能会得到改进。

关键是在单词列表中找到'and'的出现，然后用复合词替换'and'及其邻居（将邻居与'and'连接起来）并将其添加回列表，连同 'and'.

的副本

我还将 'wanted' 字符串转换为列表，以将 'fish and chips' 字符串作为不同的项目处理。

import re
from collections import Counter

# changed 'wanted' string to a list
wanted = ['fish and chips','fish','chips','steak', 'and']

cnt = Counter()

words = re.findall('\w+', open('ngh.txt').read().lower())

for word in words:

    # look for 'and', replace it and neighbours with 'comp_word'
    # slice, concatenate, and append to make new words list

    if word == 'and':
        and_pos = words.index('and')
        comp_word = str(words[and_pos-1]) + ' and '  +str(words[and_pos+1])
        words = words[:and_pos-1] + words[and_pos+2:]
        words.append(comp_word)
        words.append('and')

for word in words:
    if word in wanted:
        cnt[word] += 1

print cnt

您的文本输出将是：

Counter({'fish':2, 'and':1, 'steak':1, 'chips':1, 'fish and chips':1})

如上面的评论所述，不清楚为什么在您的理想输出中 want/expect 鱼的输出为 2，薯条的输出为 2，炸鱼薯条的输出为 1。我假设这是一个错字，因为上面的输出有 'chips':1

Answer 2

我建议使用两种适用于任何模式和任何文件的算法。第一个算法的运行时间与（文件中的字符数）* 模式数成正比。

1> 对于每个模式，搜索所有模式并创建超级模式列表。这可以通过将一种模式（例如 'cat'）与要搜索的所有模式进行匹配来完成。

patterns = ['cat', 'cat and dogs', 'cat and fish']
superpattern['cat']  = ['cat and dogs', 'cat and fish']

2> 在文件中搜索 'cat'，假设结果是 cat_count 3> 现在在文件中搜索 'cat' 的每个晚餐模式并得到它们的计数

for (sp  in superpattern['cat']) :
    sp_count = match sp in file.
    cat_count = cat_count - sp

这是蛮力的通用解决方案。如果我们在 Trie 中安排模式，应该能够提出线性时间解决方案。根-->f-->i-->s-->h-->a 等等。现在，当你在鱼的 h 处，而你没有得到 a 时，增加 fish_count 并转到 root。如果你得到 'a' 继续。任何时候你得到一些意想不到的东西，增加最近找到的模式的计数并转到根或转到某个其他节点（最长匹配前缀是该其他节点的后缀）。这是 Aho-Corasick 算法，您可以在维基百科或以下位置查找它： http://www.cs.uku.fi/~kilpelai/BSA05/lectures/slides04.pdf

此解决方案与文件中的字符数成线性关系。

Answer 3

定义:

想要的项目: 正在文本中搜索的字符串。

要计算想要的项目，而不是在较长的想要的项目中重新计算它们，首先计算每个项目在字符串中出现的次数。接下来，从最长到最短的顺序浏览所需物品，当您遇到较长物品中出现的较小物品时，从较短物品中减去较长物品的结果数。例如，假设您想要的项目是 "a"、"a b" 和 "a b c"，而您的文本是 "a/a/a b/a b c"。搜索每个单独的结果：{ "a": 4, "a b": 2, "a b c": 1 }。期望的结果是：{ "a b c": 1, "a b": #("a b") - #("a b c") = 2 - 1 = 1, "a": #("a") - #("a b c") - #("a b") = 4 - 1 - 1 = 2}.

def get_word_counts(text, wanted):
    counts = {}; # The number of times a wanted item was read

    # Dictionary mapping word lengths onto wanted items
    #  (in the form of a dictionary where keys are wanted items)
    lengths = {}; 

    # Find the number of times each wanted item occurs
    for item in wanted:
        matches = re.findall('\b' + item + '\b', text);

        counts[item] = len(matches)

        l = len(item) # Length of wanted item

        # No wanted item of the same length has been encountered
        if (l not in lengths):
            # Create new dictionary of items of the given length
            lengths[l] = {}

        # Add wanted item to dictionary of items with the given length
        lengths[l][item] = 1

    # Get and sort lenths of wanted items from largest to smallest
    keys = lengths.keys()
    keys.sort(reverse=True)

    # Remove overlapping wanted items from the counts working from
    #  largest strings to smallest strings
    for i in range(1,len(keys)):
        for j in range(0,i):
            for i_item in lengths[keys[i]]:
                for j_item in lengths[keys[j]]:
                    #print str(i)+','+str(j)+': '+i_item+' , '+j_item
                    matches = re.findall('\b' + i_item + '\b', j_item);

                    counts[i_item] -= len(matches) * counts[j_item]

    return counts

以下代码包含测试用例：

tests = [
    {
        'text': 'test file with words fish, steak fish chips fish and '+
            'chips and fries',
        'wanted': ["fish and chips","fish","chips","steak"]
    },
    {
        'text': 'fish, fish and chips, fish and chips and burgers',
        'wanted': ["fish and chips","fish","fish and chips and burgers"]
    },
    {
        'text': 'fish, fish and chips and burgers',
        'wanted': ["fish and chips","fish","fish and chips and burgers"]
    },
    {
        'text': 'My fish and chips and burgers. My fish and chips and '+
            'burgers',
        'wanted': ["fish and chips","fish","fish and chips and burgers"]
    },
    {
        'text': 'fish fish fish',
        'wanted': ["fish fish","fish"]
    },
    {
        'text': 'fish fish fish',
        'wanted': ["fish fish","fish","fish fish fish"]
    }
]

for i in range(0,len(tests)):
    test = tests[i]['text']
    print test
    print get_word_counts(test, tests[i]['wanted'])
    print ''

输出结果如下：

test file with words fish, steak fish chips fish and chips and fries
{'fish and chips': 1, 'steak': 1, 'chips': 1, 'fish': 2}

fish, fish and chips, fish and chips and burgers
{'fish and chips': 1, 'fish and chips and burgers': 1, 'fish': 1}

fish, fish and chips and burgers
{'fish and chips': 0, 'fish and chips and burgers': 1, 'fish': 1}

My fish and chips and burgers. My fish and chips and burgers
{'fish and chips': 0, 'fish and chips and burgers': 2, 'fish': 0}

fish fish fish
{'fish fish': 1, 'fish': 1}

fish fish fish
{'fish fish fish': 1, 'fish fish': 0, 'fish': 0}

计算文件中的字符串，一些单个单词，一些完整的句子

Count strings in a file, some single words, some full sentences

text

counter

word-count

python-2.7