This project has moved to Flash Text <https://github.com/vi3k6i5/flashtext>_.

synonym-extractor

Synonym Extractor is a python library that is loosely based on Aho-Corasick algorithm <https://en.wikipedia.org/wiki/Aho%E2%80%93Corasick_algorithm>_.

The idea is to extract words that we care about from a given sentence in one pass.

Basically say I have a vocabulary of 10K words and I want to get all the words from that set present in a sentence. A simple regex match will take a lot of time to loop over the 10K documents.

Hence we use a simpler yet much faster algorithm to get the desired result.

Installation

::

pip install synonym-extractor

Usage

::

# import module
from synonym.extractor import SynonymExtractor

# Create an object of SynonymExtractor
synonym_extractor = SynonymExtractor()

# add synonyms
synonym_names = ['NY', 'new-york', 'SF']
clean_names = ['new york', 'new york', 'san francisco']

for synonym_name, clean_name in zip(synonym_names, clean_names):
    synonym_extractor.add_to_synonym(synonym_name, clean_name)

synonyms_found = synonym_extractor.get_synonyms_from_sentence('I love SF and NY. new-york is the best.')

synonyms_found
>> ['san francisco', 'new york', 'new york']

Algorithm

synonym-extractor is based on Aho-Corasick algorithm <https://en.wikipedia.org/wiki/Aho%E2%80%93Corasick_algorithm>_.

Documentation

Documentation can be found at Read the Docs <http://synonym-extractor.readthedocs.org>_.

Why

::

Say you have a corpus where similar words appear frequently.

eg: Last weekened I was in NY. I am traveling to new york next weekend.

If you train a word2vec model on this or do any sort of NLP it will treat NY and new york as 2 different words.

Instead if you create a synonym dictionary like:

eg: NY=>new york new york=>new york

Then you can extract NY and new york as the same text.

To do the same with regex it will take a lot of time:

============ ========== = ========= ============ Docs count # Synonyms : Regex synonym-extractor ============ ========== = ========= ============ 1.5 million 2K : 16 hours NA 2.5 million 10K : 15 days 15 mins ============ ========== = ========= ============

The idea for this library came from the following StackOverflow question <https://stackoverflow.com/questions/44178449/regex-replace-is-taking-time-for-millions-of-documents-how-to-make-it-faster>_.

License

The project is licensed under the MIT license.