Manual Spell Checker

A manual spell checker built on pyenchant that allows you to swiftly correct misspelled words.

Why does this exist?

While I was working on a text based multi-class classification competition, I noticed that the data contained a lot of misspelled words, errors which automated spell check packages out there couldn't fix. This was because the data had been compiled based on a survey of people who weren't native English speakers. As the there weren't many samples in the dataset (~1000), I decided to write some code for automated detection of spelling errors which I could then fix manually, and thus, this package was born.

How to install?

pip install manual_spellchecker

Features

• All features as provided by pyenchant
• Quickly analyze and get a list of all misspelled words
• Can replace, skip and delete misspelled words
• Use your favourite tokenizer for splitting words
• Replaced misspelled words via provided suggestions by simply typing in their indices (Indexing starts from 1 and not 0)
• Can checkpoint current set of corrections
• Contexualized pretty printing for easy visual correction (works on both command line and notebook)

Functions and Parameters

# Initialize the spell checking object
__init__(dataframe, column_names, tokenizer=None, num_n_words_dis=5, save_path=None)

• dataframe - Takes a pandas dataframe as input
• column_names - Pass the column name(s) upon which you want to perform spelling correction
• tokenizer=None - Pass your favourite tokenizer like nltk or spacy, etc. (Default: splits on space)
• num_n_words_dis=5 - This decides how many neighbouring words to display on either side of the error
• save_path=None - If a save path is provided, the final corrected dataframe is saved as a csv. (Default: the dataframe is not saved but returned)

# For quick analysis of all the misspelled words
spell_check()

# Returns a list of all the misspelled words
get_all_errors()

# Starts the process of correcting erroneous words
correct_words()

Important Note:

• Type -999 into the input box to stop the error correction and save the current progress (if save_path is provided)
• Simply press enter if you want to skip the current word
• Type in "" or '' in the input box to delete a misspelled word

Usage

How to import?

from manual_spellchecker import spell_checker

Quick analysis of the total number of errors

# Read the data
df = pd.read_csv("Train.csv")
# Initialize the model
ob = spell_checker(df, "text")
# Quick analysis
ob.spell_check()

Multiple columns can be passed for spelling correction

# Read the data
df = pd.read_csv("Train.csv")
# Initialize the model
ob = spell_checker(df, ["text", "label"])
# Quick analysis
ob.spell_check()

You can pass your own tokenizers

# Import nltk's word tokenizer
from nltk import word_tokenize
# Read the data
df = pd.read_csv("Train.csv")
# Initialize the model
ob = spell_checker(df, "text", word_tokenize)
# Quick analysis
ob.spell_check()

Get a list of all the errors

# Read the data
df = pd.read_csv("Train.csv")
# Initialize the model
ob = spell_checker(df, "text")
# Quick analysis. This needs to be performed before getting all errors
ob.spell_check()
# Returns a list of all errors
ob.get_all_errors()

Make corrections

# Read the data
df = pd.read_csv("Train.csv")
# Initialize the model
ob = spell_checker(df, "text")
# Start corrections
ob.correct_words()

To save

df = pd.read_csv("Train.csv")
# Initialize the model
ob = spell_checker(df, "text", save_path="correct_train_data.csv")

Future Ideas

• Will be adding automated, contextual error corrections

Feature Request

Drop me an email at atif.hit.hassan@gmail.com if you want any particular feature