madhav-datt / autocorrect-lib

Autocorrect Library - probabilistic model based word spelling corrections

autocorrect-lib

Autocorrect Library - probabilistic model based word spelling corrections. C language Dictionary Library designed to use the English dictionary in Trie data structures for fast autocorrection of incorrectly spelled words.

Library Contents and Functions

• int AUTOCORR_LENGTH_MAX - Maximum permitted word length that can be checked and corrected using the library.
• int AUTOCORR_check_word (const char* word) - Check if word is present in word data trie. Returns word_frequency if word is in present in trie, else returns -1.
• bool AUTOCORR_upload (void) - Load word frequency data from file into a Trie data structure. Returns true if successful else false. To be called only once before any other functions from the library can be used.
• char* AUTOCORR_correct_word (const char* word) - Return word with highest probability value as per words_hash_table as suggested correction for misspelled word. To be called for every misspelled word in file.
• int AUTOCORR_size_data (void) - Return size of word data file, that is, number of words in the frequency data file.
• void AUTOCORR_unload_words (void) - Unload word frequency data from memory. None of the above functions can be called once the file has been unloaded.

Probability based Correction

char* AUTOCORR_correct_word (const char* word) finds words with edit distance = 1 and returns word with highest probability value as per word_data.txt as suggested correction for the misspelled word.

Edit distance between two words is defined as the number of edits to turn one into the other. Can be a deletion (remove one letter), a transposition (swap adjacent letters), an alteration (change one letter to another) or an insertion (add a letter).

Word Segmentation based Correction

If finding correct words within the acceptable edit distance value fails, char* AUTOCORR_correct_word (const char* word) tries to break the down by inserting spaces.

A greedy algorithm is used to select the longest correct word prefix of the the word being corrected. The algorithm is recursively run on the suffix till all the missing spaces have been added.

The last of the misspelled portion (no more spaces can be added to obtain correctly spelled words) is discarded.

Using the library

Downloading the library

$ wget ~/Downloads https://github.com/madhav-datt/autocorrect-lib/archive/autocorrect-lib.v1.0.zip
$ unzip autocorrect-lib.v1.0.zip
$ rm -f autocorrect-lib.v1.0.zip

Using the library in a program

Include the following header file in your program:

#include "/Downloads/autocorrect-lib.v1.0/lib/autocorrect.h"

TODOs

• Autocorrect error correction probability improvement.

  • Take into account words with 2 corrections instead of 1.
  • Consider word hash values with frequency > 1 and Edit Distance = 2.

• Learning from user inputs

  • Add text from files checked by user into word data.
  • Process and update word frequency file.

• Optimize library for space. Currently, consumes a lot of memory.

• Correct words with missing spaces.