SymSpellJPy

This is a python wrapper module for a Java implementation of the SymSpell library.

Dependencies

1. python3.6: conda create --name <ENV_NAME> python=3.6
2. Java 1.8 SDK

Install

1. Install Dependencies
2. Activate the Python virtual environment: conda activate <ENV_NAME>
3. Install SymSpellJPy: pip install symspelljpy

Usage

import symspelljpy

spell_client = symspelljpy.SymSpellClient(distance_type='QWE')
print(spell_client.lookup('plase correcme'))

{"inputText":"plase correcme","output":[{"outputText":"please correct me ","mode":"COMPOUND","distance":3.0,"count":4.4467949E7}]}

This python wrapper is build on top of the following jar file:

$ java -jar ./symspell-console/target/spellcheckclient-jar-with-dependencies.jar -h
usage: java -jar
            symspell-console-6.6-SNAPSHOT-jar-with-dependencies.jar.jar
 -b,--bigram <arg>     bi-gram dictionary file path
 -d,--distance <arg>   spelling correction distance type:
                       'VDL': vanilla Damerau Levenshtein distance.
                       'WDL': weighted Damerau Levenshtein distance.
                       'QWE': qwerty distance.
 -e,--edits <arg>      maximum number of edits (default 2)
 -h,--help             this help message
 -k,--topk <arg>       number of candidates to output (default 5)
 -m,--mode <arg>       spelling correction mode: 'SMART'(Default), 'ALL',
                       'WORD', 'COMPOUND' or 'SEGMENTATION'.
                       WORD: Individual word spelling correction.
                       COMPOUND: Compound splitting/decompounding +
                       Automatic spelling correction. Space can only be
                       inserted/deleted for a token once.
                       SEGMENTATION: Word segmentation  + Automatic
                       spelling correction. Existing spaces are allowed
                       and considered for optimum segmentation.
                       SMART: when there is no space in the input text and
                       the text length is over the maximum word length,
                       enable word segmenation. Otherwise choose COMPOUND
                       word correction model.
                       ALL: COMPOUND + SEGMENTATION.
 -t,--timer            execution time per input in milliseconds.
 -u,--unigram <arg>    uni-gram dictionary file path
 -w,--word <arg>       maximum word length for word segmentation (default
                       10)