Date spaCy

Date spaCy is a collection of custom spaCy pipeline component that enables you to easily identify date entities in a text and fetch the parsed date values using spaCy's token extensions. It uses RegEx to find dates and then uses the dateparser library to convert those dates into structured datetime data. One current limitation is that if no year is given, it presumes it is the current year. The dateparser output is stored in a custom entity extension: ._.date.

This lightweight approach can be added to an existing spaCy pipeline or to a blank model. If using in an existing spaCy pipeline, be sure to add it before the NER model.

Installation

To install date_spacy_pt, simply run:

pip install git+https://github.com/BornSaint/date_spacy_pt.git

Usage

Adding the Component to your spaCy Pipeline

First, you'll need to import the find_dates component and add it to your spaCy pipeline:

import spacy
from date_spacy import find_dates

# Load your desired spaCy model
nlp = spacy.blank('pt')

# Add the component to the pipeline
nlp.add_pipe('find_dates')

Processing Text with the Pipeline

After adding the component, you can process text as usual:

doc = nlp('''o evento está agendado pro dia 25 de Agosto de 2023.
  Nós também temos uma reunião para 10 de setembro e uma outra no décimo segundo dia de outubro de 2024
  décimo-segundo dia de outubro de 2024
  e finaliza no dia 4 de Jan''')

Accessing the Parsed Dates

You can iterate over the entities in the doc and access the special date extension:

for ent in doc.ents:
    if ent.label_ == "DATE":
        print(f"Text: {ent.text} -> Parsed Date: {ent._.date}")

This will output:

Text: 25 de Agosto de 2023 -> Parsed Date: 2023-08-25 00:00:00
Text: 10 de setembro -> Parsed Date: 2024-09-10 00:00:00
Text: décimo segundo dia de outubro de 2024 -> Parsed Date: 2024-10-12 00:00:00
Text: décimo-segundo dia de outubro de 2024 -> Parsed Date: 2024-10-12 00:00:00
Text: 4 de Jan -> Parsed Date: 2024-01-04 00:00:00