這幾天在翻舊程式碼時發現以前寫的註釋部分有很多單詞拼寫錯誤,這些單詞錯得不算離譜,應該可以用工具自動糾錯絕大部分。用 Python 寫個拼寫檢查指令碼很容易,如果能很好利用 aspell/ispell 這些現成的小工具就更簡單了。
要點
1、輸入一個拼寫錯誤的單詞,呼叫 aspell -a 後得到一些候選正確單詞,然後用距離編輯進一步嗮選出更精確的詞。比如執行 aspell -a,輸入 ‘hella’ 後得到如下結果:
hell, Helli, hello, heal, Heall, he’ll, hells, Heller, Ella, Hall, Hill, Hull, hall, heel, hill, hula, hull, Helga, Helsa, Bella, Della, Mella, Sella, fella, Halli, Hally, Hilly, Holli, Holly, hallo, hilly, holly, hullo, Hell’s, hell’s
2、什麼是距離編輯(Edit-Distance,也叫 Levenshtein algorithm)呢?就是說給定一個單詞,通過多次插入、刪除、交換、替換單字元的操作後列舉出所有可能的正確拼寫,比如輸入 ‘hella’,經過多次插入、刪除、交換、替換單字元的操作後變成:
‘helkla’, ‘hjlla’, ‘hylla’, ‘hellma’, ‘khella’, ‘iella’, ‘helhla’, ‘hellag’, ‘hela’, ‘vhella’, ‘hhella’, ‘hell’, ‘heglla’, ‘hvlla’, ‘hellaa’, ‘ghella’, ‘hellar’, ‘heslla’, ‘lhella’, ‘helpa’, ‘hello’, …
3、綜合上面2個集合的結果,並且考慮到一些理論知識可以提高拼寫檢查的準確度,比如一般來說寫錯單詞都是無意的或者誤打,完全錯的單詞可能性很小,而且單詞的第一個字母一般不會拼錯。所以可以在上面集合裡去掉第一個字母不符合的單詞,比如:’Sella’, ‘Mella’, khella’, ‘iella’ 等,這裡 VPSee 不刪除單詞,而把這些單詞從佇列裡取出來放到佇列最後(優先順序降低),所以實在匹配不了以 h 開頭的單詞才去匹配那些以其他字母開頭的單詞。
4、程式中用到了外部工具 aspell,如何在 Python 裡捕捉外部程式的輸入和輸出以便在 Python 程式裡處理這些輸入和輸出呢?Python 2.4 以後引入了 subprocess 模組,可以用 subprocess.Popen 來處理。
5、Google 大牛 Peter Norvig 寫了一篇 How to Write a Spelling Corrector 很值得一看,大牛就是大牛,21行 Python 就解決拼寫問題,而且還不用外部工具,只需要事先讀入一個詞典檔案。本文程式的 edits1 函式就是從牛人家那裡 copy 的。
程式碼
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 |
#!/usr/bin/python # A simple spell checker # written by http://www.vpsee.com import os, sys, subprocess, signal alphabet = 'abcdefghijklmnopqrstuvwxyz' def found(word, args, cwd = None, shell = True): child = subprocess.Popen(args, shell = shell, stdin = subprocess.PIPE, stdout = subprocess.PIPE, cwd = cwd, universal_newlines = True) child.stdout.readline() (stdout, stderr) = child.communicate(word) if ": " in stdout: # remove nn stdout = stdout.rstrip("n") # remove left part until : left, candidates = stdout.split(": ", 1) candidates = candidates.split(", ") # making an error on the first letter of a word is less # probable, so we remove those candidates and append them # to the tail of queue, make them less priority for item in candidates: if item[0] != word[0]: candidates.remove(item) candidates.append(item) return candidates else: return None # copy from http://norvig.com/spell-correct.html def edits1(word): n = len(word) return set([word[0:i]+word[i+1:] for i in range(n)] + [word[0:i]+word[i+1]+word[i]+word[i+2:] for i in range(n-1)] + [word[0:i]+c+word[i+1:] for i in range(n) for c in alphabet] + [word[0:i]+c+word[i:] for i in range(n+1) for c in alphabet]) def correct(word): candidates1 = found(word, 'aspell -a') if not candidates1: print "no suggestion" return candidates2 = edits1(word) candidates = [] for word in candidates1: if word in candidates2: candidates.append(word) if not candidates: print "suggestion: %s" % candidates1[0] else: print "suggestion: %s" % max(candidates) def signal_handler(signal, frame): sys.exit(0) if __name__ == '__main__': signal.signal(signal.SIGINT, signal_handler) while True: input = raw_input() correct(input) |
更簡單的方法
當然直接在程式裡呼叫相關模組最簡單了,有個叫做 PyEnchant 的庫支援拼寫檢查,安裝 PyEnchant 和 Enchant 後就可以直接在 Python 程式裡 import 了:
1 2 3 4 5 6 7 8 9 |
>>> import enchant >>> d = enchant.Dict("en_US") >>> d.check("Hello") True >>> d.check("Helo") False >>> d.suggest("Helo") ['He lo', 'He-lo', 'Hello', 'Helot', 'Help', 'Halo', 'Hell', 'Held', 'Helm', 'Hero', "He'll"] >>> |