Pandas 中的拼写检查器

Question

我正在尝试使用从 SQL 数据库中提取的单词在 pandas 类中实现 Peter Norvig's spell checker。数据包含用户查询，这些查询通常包含许多拼写错误，我希望这个类将返回最有可能的查询（拼写正确）。

该类使用返回 pandas 数据帧的数据库查询进行初始化。例如：

  query     count
0 foo bar       1864
1 super foo      73
2 bar of foos    1629
3 crazy foos     940

以下大部分内容是直接从 Peter 的工作中提取的，但我对课程所做的修改似乎无法正常工作。我的猜测是它与删除计数器功能有关 (WORDS = Counter(words(open('big.txt').read())))，但我不确定从数据帧中获取相同功能的最佳方法。

当前课程如下：

class _SpellCheckClient(object):
  """Wraps functionality to check the spelling of a query."""

  def __init__(self, team, table, dremel_connection):
    self.df = database_connection.ExecuteQuery(
        'SELECT query, COUNT(query) AS count FROM table GROUP BY 1;' 

  def expected_word(self, word):
    """Most probable spelling correction for word."""
    return max(self._candidates(word), key=self._probability)

  def _probability(self, query):
    """Probability of a given word within a query."""
    query_count = self.df.loc[self.df['query'] == query]['count'].values
    return query_count / self.df['count'].sum()

  def _candidates(self, word):
    """Generate possible spelling corrections for word."""
    return (self._known([word])
            or self._known(self._one_edits_from_word(word))
            or self._known(self._two_edits_from_word(word))
            or [word])

  def _known(self, query):
    """The subset of `words` that appear in the dictionary of WORDS."""
    # return set(w for w in query if w in WORDS)
    return set(w for w in query if w in self.df['query'].value_counts)

  def _one_edits_from_word(self, word):
    """All edits that are one edit away from `word`."""
    splits = [(word[:i], word[i:]) for i in xrange(len(word) + 1)]
    deletes = [left + right[1:] for left, right in splits if right]
    transposes = [left + right[1] + right[0] + right[2:]
                  for left, right in splits
                  if len(right) > 1]
    replaces = [left + center + right[1:]
                for left, right in splits
                if right for center in LETTERS]
    inserts = [left + center + right
               for left, right in splits
               for center in LETTERS]
    return set(deletes + transposes + replaces + inserts)

  def _two_edits_from_word(self, word):
    """All edits that are two edits away from `word`."""
    return (e2 for e1 in self._one_edits_from_word(word)
            for e2 in self._one_edits_from_word(e1))

提前致谢！

Answer 1

import pandas as pd
from spellchecker import SpellChecker
df = pd.Series(['Customir','Tast','Hlp'])
spell = SpellChecker(distance=1)
def Correct(x):
    return spell.correction(x)
df = df.apply(Correct)
df

0    customer
1        last
2        help
dtype: object

Answer 2

对于正在寻找答案的任何人，以下是对我有用的方法：

def _words(df):
  """Returns the total count of each word within a dataframe."""
  return df['query'].str.get_dummies(sep=' ').T.dot(df['count'])


class _SpellCheckClient(object):
  """Wraps functionality to check the spelling of a query."""

  def __init__(self, team, table, database_connection):
    self.df = database_connection
    self.words = _words(self.df)

  def expected_word(self, query):
    """Most probable spelling correction for word."""
    return max(self._candidates(query), key=self._probability)

  def _probability(self, query):
    """Probability of a given word within a query."""
    return self.words.pipe(lambda x: x / x.sum()).get(query, 0.0)

  def _candidates(self, query):
    """Generate possible spelling corrections for word."""
    return (self._known(self._one_edits_from_query(query))
            or self._known(self._two_edits_from_query(query))
            or [query])

  def _known(self, query):
    """The subset of `query` that appear in the search console database."""
    return set(w for w in query if self.words.get(w))

  def _one_edits_from_query(self, query):
    """All edits that are one edit away from `query`."""
    splits = [(query[:i], query[i:]) for i in xrange(len(query) + 1)]
    deletes = [left + right[1:] for left, right in splits if right]
    transposes = [left + right[1] + right[0] + right[2:]
                  for left, right in splits
                  if len(right) > 1]
    replaces = [left + center + right[1:]
                for left, right in splits
                if right for center in LETTERS]
    inserts = [left + center + right
               for left, right in splits
               for center in LETTERS]
    return set(deletes + transposes + replaces + inserts)

  def _two_edits_from_query(self, query):
    """All edits that are two edits away from `query`."""
    return (e2 for e1 in self._one_edits_from_query(query)
            for e2 in self._one_edits_from_query(e1))