autocorrect_package/autocorrection.py

# Import necessary libraries
import string  # For lowercase English letters
import re      # For regular expression operations
from collections import Counter  # For counting word occurrences
import editdistance  # For calculating edit distance
import numpy as np  # Not used in code (can be removed)

class Autocorrection(object):

    def __init__(self, filename):
        # Read the file and extract all lowercase words
        with open(filename, "r", encoding="UTF-8") as file:
            word = []
            lines = file.readlines()
            for line in lines:
                # Extract words using regex and convert to lowercase
                word += re.findall(r'\w+', line.lower())

        # Create a vocabulary from the unique words
        self.vocabulary = set(word)

        # Count occurrences of each word
        self.counts_of_word = Counter(word)

        # Total number of words for probability calculation
        self.total_words = float(sum(self.counts_of_word.values()))

        # Compute the probability of each word
        self.prob_of_word = {
            w: self.counts_of_word[w] / self.total_words
            for w in self.counts_of_word.keys()
        }

    def edit1(self, word):
        # All letters from a-z
        # letter = string.ascii_lowercase
        letter = 'abcdefghijklmnopqrstuvwxyzàâçéèêëîïôùûüÿœæ'

        # All possible split positions in the word
        splits = [(word[:i], word[i:]) for i in range(len(word) + 1)]

        # Generate all one-edit-distance words:
        # Insertion: insert a letter at any position
        insert = [l + c + r for l, r in splits for c in letter]

        # Deletion: remove one letter
        delete = [l + r[1:] for l, r in splits if r]

        # Replacement: change one letter
        replace = [l + c + r[1:] for l, r in splits if r for c in letter]

        # Transposition (swap): swap adjacent letters
        swap = [l + r[1] + r[0] + r[2:] for l, r in splits if len(r) > 1]

        # Return all variations as a set to avoid duplicates
        return set(replace + insert + delete + swap)

    def edit2(self, word):
        # Generate all words that are two edits away
        return [e2 for e1 in self.edit1(word) for e2 in self.edit1(e1)]

    def common_prefix_length(self, word1, word2):
        # Find the length of the common prefix between two words
        common_len = 0
        for i, (c1, c2) in enumerate(zip(word1, word2)):
            if c1 == c2:
                common_len += 1
            else:
                break
        return common_len

    def custom_score(self, suggestion, original_word):
        # Weights for different operations
        replace_weight = 1
        insert_weight = 2
        delete_weight = 3
        swap_weight = 4

        # Get the edit distance between suggestion and original word
        distance = editdistance.eval(suggestion, original_word)

        # Compute a custom score based on the type of edit
        if len(suggestion) == len(original_word):  # Likely replace
            common_prefix_len = self.common_prefix_length(suggestion, original_word)
            score = distance * replace_weight + (len(original_word) - common_prefix_len)
        elif len(suggestion) == len(original_word) + 1:  # Likely insertion
            common_prefix_len = self.common_prefix_length(suggestion, original_word)
            score = distance * insert_weight + (len(original_word) - common_prefix_len)
        elif len(suggestion) == len(original_word) - 1:  # Likely deletion
            score = distance * delete_weight
        else:  # Likely swap or other
            score = distance * swap_weight

        return score

    def correct_spelling(self, word):
        # Return if word is already in the vocabulary
        if word in self.vocabulary:
            print(f"{word} is already correctly spelt")
            return

        # Get edit distance candidates
        suggestions = self.edit1(word) or self.edit2(word) or [word]

        # Filter candidates that are real words
        best_guesses = [w for w in suggestions if w in self.vocabulary]

        # Sort suggestions by custom score
        best_guesses.sort(key=lambda w: self.custom_score(w, word))

        # Return suggestions with their probabilities
        return [(w, self.prob_of_word[w]) for w in best_guesses]