chunk english

Introduction

The English chunking model in Flair is a phrase chunking model designed for token classification tasks. It uses Flair embeddings combined with an LSTM-CRF architecture to predict tags such as noun phrases (NP), verb phrases (VP), prepositional phrases (PP), and more. The model achieves an F1-Score of 96.48 on the CoNLL-2000 dataset.

Architecture

The model is based on Flair embeddings and LSTM-CRF architecture. It utilizes contextual string embeddings both forward and backward, and is designed to recognize and label different kinds of phrases in English text.

Training

The model is trained using a script in Flair with the CoNLL-2000 dataset. It employs stacked embeddings, including Flair and GloVe embeddings, and trains a sequence tagger to recognize noun phrases. The training involves initializing a tag dictionary, embedding types, and a sequence tagger, followed by training the model over 150 epochs.

Guide: Running Locally

1. Install Flair: Install the Flair library using pip:

   pip install flair
   

2. Load the Tagger: Use the Flair library to load the sequence tagger for English chunking:

   from flair.data import Sentence
   from flair.models import SequenceTagger
   
   tagger = SequenceTagger.load("flair/chunk-english")
   

3. Make Predictions: Create a sentence, predict the tags, and print the results:

   sentence = Sentence("The happy man has been eating at the diner")
   tagger.predict(sentence)
   print(sentence)
   for entity in sentence.get_spans('np'):
       print(entity)
   

4. Suggestions for Cloud GPUs: For efficient training and inference, consider using cloud GPU services such as AWS, Google Cloud, or Azure.

License

The model and code are subject to the licensing terms outlined in the Flair project's repository. Users should refer to the Flair GitHub repository for detailed licensing information.