PatternOmatic 0.2.*

#AI · #EvolutionaryComputation · #NLP

Discover spaCy's linguistic patterns matching a given set of string samples

Requirements

• Python 3.7.3
• Spacy 2.3.*

Basic usage

From sources

Clone SCM official repository

git clone git@github.com:revuel/PatternOmatic.git

Play with Makefile

• make venv to activate project's Virtual Environment*
• make libs to install dependencies
• make test to run Unit Tests
• make coverage to run Code Coverage
• make run to run PatternOmatic's script with example parameters

_{* you must have one first}

From package

Install package

pip install PatternOmatic

Play with the CLI

# Show help 
patternomatic.py -h

# Usage example 1: Basic
patternomatic.py -s Hello world -s Goodbye world

# Usage example 2: Using a different language
python -m spacy download es_core_news_sm
patternomatic.py -s Me llamo Miguel -s Se llama PatternOmatic -l es_core_news_sm

Play with the library

""" 
PatternOmatic library client example.
Find linguistic patterns to be used by the spaCy Rule Based Matcher

"""
from PatternOmatic.api import find_patterns, Config

if __name__ == '__main__':

    my_samples = ['I am a cat!', 'You are a dog!', 'She is an owl!']

    # Optionally, let it evolve a little bit more!
    config = Config()
    config.max_generations = 150
    config.max_runs = 3

    patterns_found, _ = find_patterns(my_samples)

    print(f'Patterns found: {patterns_found}')

---

Features

Generic

✅ No OS dependencies, no storage or database required!

✅ Lightweight package with just a little direct pip dependencies

• spaCy
• spaCy's en_core_web_sm Language Model

✅ Easy and highly configurable to boost clever searches

✅ Includes basic logging mechanism

✅ Includes basic reporting, JSON and CSV format supported. Report file path is configurable

✅ Configuration file example provided (config.ini)

✅ Default configuration is run if no configuration file provided

✅ Provides rollback actions against several possible misconfiguration scenarios

Evolutionary

✅ Basic Evolutionary (Grammatical Evolution) parameters available and configurable

✅ Supports two different Evolutionary Fitness functions

✅ Supports Binary Tournament Evolutionary Selection Type

✅ Supports Random One Point Crossover Evolutionary Recombination Type

✅ Supports "µ + λ" Evolutionary Replacement Type

✅ Supports "µ ∪ λ" with elitism Evolutionary Replacement Type

✅ Supports "µ ∪ λ" without elitism Evolutionary Replacement Type

✅ Typical evolutionary performance metrics included:

• Success Rate (SR)
• Mean Best Fitness (MBF)
• Average Evaluations to Solution (AES)

Linguistic

✅ Compatible with any spaCy Language Model

✅ Supports all spaCy's Rule Based Matcher standard Token attributes

✅ Supports the following spaCy's Rule Based Matcher non standard Token attributes (via underscore)

• ent_id
• ent_iob
• ent_kb_id
• has_vector
• is_bracket
• is_currency
• is_left_punct
• is_oov
• is_quote
• is_right_punct
• lang
• norm
• prefix
• sentiment
• string
• suffix
• text_with_ws
• whitespace

✅ Supports skipping boolean Token attributes

✅ Supports spaCy's Rule Based Matcher Extended Pattern Syntax

✅ Supports spaCy's Rule Based Matcher Grammar Operators and Quantifiers

✅ Supports Token Wildcard

✅ Supports defining the number of attributes per token within searched patterns

✅ Supports usage of non repeated token attribute values

---

Author: Miguel Revuelta Espinosa (revuel), a humble AI enthusiastic