NLP预处理阶段----拼写纠错实战

拼写纠错流程

仅限拼写上的纠错，暂不去进行语法层次上的纠错。
I like play football. 本文暂不纠错。

数据集：

spell-errors.txt

正确：错误1，错误2… 其他错误给予小概率
vocab.txt 词典

testdata.txt 测试数据

编号–错误量–文本

流程图

1.构建词典库

# 词典库
# 转换成集合复杂度O(logn),列表复杂度为O(n)
vocab = set([line.rstrip() for line in open('vocab.txt')])
print(vocab)

2.基于词典库生成编辑距离不大于2的近似词

# 需要生成所有候选集合
def generate_candidates(word):
    """
    word: 给定的输入（错误的输入） 
    返回所有(valid)候选集合
    """
    # 生成编辑距离为1的单词
    # 1.insert 2. delete 3. replace
    # appl: replace: bppl, cppl, aapl, abpl... 
    #       insert: bappl, cappl, abppl, acppl....
    #       delete: ppl, apl, app
    
    # 假设使用26个字符
    letters = 'abcdefghijklmnopqrstuvwxyz' 
    
    splits = [(word[:i], word[i:]) for i in range(len(word)+1)]
    # insert操作
    inserts = [L+c+R for L, R in splits for c in letters]
    # delete
    deletes = [L+R[1:] for L,R in splits if R]
    # replace
    replaces = [L+c+R[1:] for L,R in splits if R for c in letters]    
    candidates = set(inserts+deletes+replaces)    
    # 过来掉不存在于词典库里面的单词
    return [word for word in candidates if word in vocab]     
print(generate_candidates("apple"))
['apple', 'apples', 'ample', 'apply']

3.读取语料库用于建立语言模型

from nltk.corpus import reuters
# 读取语料库
categories = reuters.categories()
corpus = reuters.sents(categories=categories)
print(corpus)
[['ASIAN', 'EXPORTERS', 'FEAR', 'DAMAGE', 'FROM', 'U', '.', 'S', '.-', 'JAPAN', 'RIFT', 'Mounting', 'trade', 'friction', 'between', 'the', 'U', '.', 'S', '.', 'And', 'Japan', 'has', 'raised', 'fears', 'among', 'many', 'of', 'Asia', "'", 's', 'exporting', 'nations', 'that', 'the', 'row', 'could', 'inflict', 'far', '-', 'reaching', 'economic', 'damage', ',', 'businessmen', 'and', 'officials', 'said', '.'], ['They', 'told', 'Reuter', 'correspondents', 'in', 'Asian', 'capitals', 'a', 'U', '.', 'S', '.', 'Move', 'against', 'Japan', 'might', 'boost', 'protectionist', 'sentiment', 'in', 'the', 'U', '.', 'S', '.', 'And', 'lead', 'to', 'curbs', 'on', 'American', 'imports', 'of', 'their', 'products', '.'], ...]

4.基于语料库构建bigram语言模型的统计字典

# 构建语言模型: bigram
term_count = {}
bigram_count = {}
for doc in corpus:
    doc = ['<s>'] + doc
    for i in range(0, len(doc)-1):
        # bigram: [i,i+1]
        term = doc[i]
        bigram = doc[i:i+2]        
        if term in term_count:
            term_count[term]+=1
        else:
            term_count[term]=1
        bigram = ' '.join(bigram)
        if bigram in bigram_count:
            bigram_count[bigram]+=1
        else:
            bigram_count[bigram]=1
print(term_count)
{'<s>': 54716, 'ASIAN': 12, 'EXPORTERS': 46, 'FEAR': 2, 'DAMAGE': 13, 'FROM': 208, 'U': 6388, '.': 45900, 'S': 6382, '.-': 167, 'JAPAN': 295, 'RIFT': 1, 'Mounting': 1, 'trade': 2271, 'friction': 28, 'between': 1053, 'the': 58251, 'And': 227, 'Japan': 1594, 'has': 4679, 'raised': 334, 'fears': 65, 'among': 229, 'many': 297, 'of': 35979, 'Asia': 71, "'": 11272, 's': 9298, 'exporting': 51, 'nations': 399, 'that': 7377, 'row': 49, 'could': 1431, 'inflict': 1,......}
print(bigram_count)
{'<s> ASIAN': 4,
 'ASIAN EXPORTERS': 1,
 'EXPORTERS FEAR': 1,
 'FEAR DAMAGE': 1,
 'DAMAGE FROM': 2,
 'FROM U': 4,
 'U .': 6350,
 '. S': 5809,
 'S .-': 120,
 '.- JAPAN': 8,
 'JAPAN RIFT': 1,
 'RIFT Mounting': 1,
 'Mounting trade': 1,
 'trade friction': 17,
 'friction between': 4,
 'between the': 237,
 'the U': 1959,....}
# sklearn里面有现成的包

5.统计错误拼写及其概率–channel probability

channel_prob = {}

for line in open('spell-errors.txt'):
    items = line.split(":")
    correct = items[0].strip()
    mistakes = [item.strip() for item in items[1].strip().split(",")]
    channel_prob[correct] = {}
    for mis in mistakes:
        channel_prob[correct][mis] = 1.0/len(mistakes)
print(channel_prob)   
{'raining': {'rainning': 0.5, 'raning': 0.5}, 'writings': {'writtings': 1.0}, 'disparagingly': {'disparingly': 1.0}, 'yellow': {'yello': 1.0}, 'four': {'forer': 0.2, 'fours': 0.2, 'fuore': 0.2, 'fore*5': 0.2, 'for*4': 0.2}, 'woods': {'woodes': 1.0}, 'hanging': {'haing': 1.0}, 'aggression': {'agression': 1.0},...}

6.对拼写错误文本进行基于语言模型的拼写纠错

V = len(term_count.keys())
file = open("testdata.txt","r")
for line in file:
   items = line.rstrip().split('\t')
   line = items[2].split()
   j = 0
   for word in line:
       if word not in vocab:
           #需要替换word成正确的单词
           #Step1: 生成所有的(valid)候选集合
           candidates = generate_candidates(word)

           # 一种方式： if candidate = [], 多生成几个candidates, 比如生成编辑距离不大于2的
           # TODO ： 根据条件生成更多的候选集合           
           if len(candidates) < 1:
               continue
           probs=[]
           # 对于每一个candidate, 计算它的score
           # score = p(correct)*p(mistake|correct)
           #       = log p(correct) + log p(mistake|correct)
           # 返回score最大的candidate                  
           for candi in candidates:
               prob = 0
               # 计算channel probability
               if candi in channel_prob and word in channel_prob[candi]:
                   prob += np.log(channel_prob[candi][word])
               else:
                   prob += np.log(0.0001)
           
           #计算语言模型的概率
               pre_word = line[j-1]+" "+candi
               if pre_word in bigram_count and line[j-1] in term_count:
                   prob += np.log((bigram_count[pre_word]+1.0)/(term_count[line[j-1]]+V))
                   
               else:
                   prob += np.log(1.0/V)
               
               if j+1 < len(line):
                   pos_word = candi + " " + line[j+1]
                   if pos_word in bigram_count and candi in term_count:
                       prob += np.log((bigram_count[pos_word] + 1.0)/(term_count[candi]+V))
                   else:
                       prob += np.log(1.0/V)
               probs.append(prob)
           
           max_idx = probs.index(max(probs))
           print(word,candidates[max_idx])          
       j +=1
       export. exports
Natonal National
exports, exports
dlrs, dlrs
remainder. remainder
statementa statement
security, security
countries. countries
political, political
negotiations, negotiations
statemant statement
said. said
Justice. Justice
ananounced announced
oil. oil
avirted averted
waters. waters
histodric historic
countries, countries
said. said
Athens, Athens
Akiman, Akiman
Ozal. Ozal
disclosed. disclosed
Zeeuw, Zeeuw
Trade's Trades
agriculture, agriculture
prirce price
supports. supports
rates, rates
cmmittee committee
said. said
can't cant
eliminati eliminate
unilaterally, unilaterally
said. said
lawmakers' lawmakers
policy, policy
daliry dairy
policy, policy
herds, herds
said. said
Wallis. Wallis
Tomorrow, Tomorrow
Lyng. Lyng
week. week
states. states
May. Mayo
transaction, transaction
stock, stock
share, share
said. said
said. said
name, named
GGHF, GGHF
reporter's reporters
aftere after
bank's banks
but, but
are. area
hand, hand
that, that
dollar, dollar
higher. higher
rates. rates
grwth growth
year. year
consaquence consequence
dollar's dollars
rises, rises
1987. 1987
30, 30
dlrs, dlrs
dlrs. dlrs
period, period
dlrs, dlrs
dlrs. dlrs
chairman, chairman
line. lined
shar share
1988. 1988
copany company
dividends. dividends
repurchases, repurchases
said. said
Medtronic's Medtronics
maket market
1988, 1988
1987. 1987
products, products
pacemakers, pacemakers
valves, valves
lasers, lasers
1995. 1995
said, said
beg, beg
past, past
Medtronic's Medtronics
paecemakers pacemakers
recalls. recalls
critixized criticized
industry, industry
pacemakers. pacemakers
leads, leads
said. said
cempany company
industry. industry
Nelson, Nelson
presidelnt president
Medtronics, Medtronics
Medtronic's Medtronics
expertise, expertise
systems. systems
Watllin Wallin
acquisitions. acquisitions
plovisions provisions
won't wont
acquisition. acquisition
responsivve responsive
1988. 1988
Activitrax, Activitrax
activity. activity
AG, AG
pacemaker. pacemakers
Batignolles, Batignolles
ltSCHN.PA, ltSCHN.PA
shares. shares
1986. 1986
bonds, bonds
30, 30
sources. sources
extension, extension
Britain, Britain
udertook undertook
later. later
deilare declare
extension. extension
tomorrow, tomorrow
delegartes delegates
session. session
Cntral Central
future, future
reported. reported
Baluchp Baluch
bales, bales
year's years
production. productions
Morlin Marlin
tariffs. tariffs
..........