QuickEntity: Named Entity Recognition Training Module

Simple is better than complex

QuickEntity is a python module designed to help you train your own Named Entity Recognition (NER) model quickly and easily. With quick NER, you can customize model your NER model by providing your own list of named entities.

• Install
• Features
• Dependencies
• Usage
  • Setting Up
  • Reading Named Entity Lists
  • Process a text with the loaded entities
  • Training the text data
  • Display the annotated text
• API Reference

Install

You can install QuickEntity by runing the following command:

pip install quickentity

Features

• Easy-to-use API for training NER models
• Ability to set language and load custom named entity lists
• Automatic saving of trained model to disk

Dependencies

QuickEntity requires:

• spacy (>= 3.5.0)
• nltk (>=3.7)

Usage

Setting Up

To use QuickEntity, you need to import the QuickEntity module

from quickentity import QuickEntity

Initialize the Quick_NER object

Then, you need to create an instance of the Quick_NER class:

phrase = "Steve played a pivotal role in the development of Apple, the company responsible for creating innovative products such as the iPad"

QE = QuickEntity(language="en", phrase=phrase, save_model=False)

The language parameter specifies the language of the text you want to train the model on (default is "en"). The phrase parameter is an exemple text phrase used to create a Doc object for training. The save_model parameter specifies whether to save the treined model to disk or not (default is True).

Reading Named Entity Lists

Before training the model, you need to load entity list using the read_json

ent_list = QE.read_json("entities.json")

The named entity list should be a JSON file with a dictionary of entities and their labels with prefix B-. Here's an example:

{
"Apple":"B-ORG",
"Steve":"B-PERSON",
"iPad":"B-PRODUCT"
}

Process a text with the loaded entities

Next, process your text data using the process_text method to obtain the list of words, spaces, and entity labels. Look how to do it:

model = QE.process_text(ent_list)

Training the text data

Once you've processed your text data, you should train the model using the train method:

QE.train(model)

Display the annotated text

Visualize the results of your model using the show method:

QE.show()

Full Example:

'''
1- install:
pip install quickentity
2- punkt package from nltk is required to tokenization:
import nltk
nltk.download('punkt')
'''

from quickentity import QuickEntity

words = """ Steve played a pivotal role in the development of Apple, 
the company responsible for creating innovative products such as the iPad."""

# config the QuickEntity, phrase is required
#language is "en" by default, 
#save_model is true by default.
QE = QuickEntity(language="en",phrase=words, save_model=True)

#load entities file in json format
ent_list = QE.read_json("ent_list.json")

# process the text data to associate entities labels
model = QE.process_text(ent_list)
# train de model
QE.train(model)

# output :
# file ./train.spacy saved on disk

# view in a jupyter-based notebook.
QE.show()

Here's the output:

API Reference

QuickEntity(language, phrase, save_model)

Create an instance of the Quick_NER class.

Parameters

• language (string): Language for the NER model. Default is "en".
• phrase (string): Example text used for training.
• save_model (bool): Whether to save the treined model to disk. Default is True

Methods

set_language(language): Set the language of the NER model.

Parameters

• language (string): Language for NER model.

Methods

read_json(file): Load named entities from a JSON file.

Parameters

• file(string): Path to JSON file containing named entities.

Methods

process_text(text): Process the entities obtained from the read_json to obtain the list of words, spaces, and entity labels.

Parameters

• ent_list (object): Object processed with read_json method.

Methods

train(model): Train the NER model using the processed training data.

Parameters

• model (object) : Object obtained from the process_text method.

Methods

show() : Visualize the results of the trained model.

Parameters

• None.

License

This project is licensed under the MIT License