README.md

metadata

tags:
  - flair
  - token-classification
  - sequence-tagger-model
language: en
datasets:
  - ontonotes
inference: false

English Verb Disambiguation in Flair (default model)

This is the standard verb disambiguation model for English that ships with Flair.

F1-Score: 89,34 (Ontonotes) - predicts Proposition Bank verb frames.

Based on Flair embeddings and LSTM-CRF.

---

Demo: How to use in Flair

Requires: Flair (pip install flair)

from flair.data import Sentence
from flair.models import SequenceTagger

# load tagger
tagger = SequenceTagger.load("flair/pos-english")

# make example sentence
sentence = Sentence("I love Berlin.")

# predict NER tags
tagger.predict(sentence)

# print sentence
print(sentence)

# print predicted NER spans
print('The following NER tags are found:')
# iterate over entities and print
for entity in sentence.get_spans('pos'):
    print(entity)

This yields the following output:

Span [1]: "I"   [− Labels: PRP (1.0)]
Span [2]: "love"   [− Labels: VBP (1.0)]
Span [3]: "Berlin"   [− Labels: NNP (0.9999)]
Span [4]: "."   [− Labels: . (1.0)]

So, the word "I" is labeled as a pronoun (PRP), "love" is labeled as a verb (VBP) and "Berlin" is labeled as a proper noun (NNP) in the sentence "TheI love Berlin".

---

Training: Script to train this model

The following Flair script was used to train this model:

from flair.data import Corpus
from flair.datasets import ColumnCorpus
from flair.embeddings import WordEmbeddings, StackedEmbeddings, FlairEmbeddings

# 1. load the corpus (Ontonotes does not ship with Flair, you need to download and reformat into a column format yourself)
corpus = ColumnCorpus(
    "resources/tasks/srl", column_format={1: "text", 11: "frame"}
)


# 2. what tag do we want to predict?
tag_type = 'frame'

# 3. make the tag dictionary from the corpus
tag_dictionary = corpus.make_tag_dictionary(tag_type=tag_type)

# 4. initialize each embedding we use
embedding_types = [

    BytePairEmbeddings("en"),

    FlairEmbeddings("news-forward-fast"),

    FlairEmbeddings("news-backward-fast"),
]

# embedding stack consists of Flair and GloVe embeddings
embeddings = StackedEmbeddings(embeddings=embedding_types)

# 5. initialize sequence tagger
from flair.models import SequenceTagger

tagger = SequenceTagger(hidden_size=256,
                        embeddings=embeddings,
                        tag_dictionary=tag_dictionary,
                        tag_type=tag_type)

# 6. initialize trainer
from flair.trainers import ModelTrainer

trainer = ModelTrainer(tagger, corpus)

# 7. run training
trainer.train('resources/taggers/frame-english',
              train_with_dev=True,
              max_epochs=150)

---

Cite

Please cite the following paper when using this model.

@inproceedings{akbik2019flair,
  title={FLAIR: An easy-to-use framework for state-of-the-art NLP},
  author={Akbik, Alan and Bergmann, Tanja and Blythe, Duncan and Rasul, Kashif and Schweter, Stefan and Vollgraf, Roland},
  booktitle={{NAACL} 2019, 2019 Conference of the North American Chapter of the Association for Computational Linguistics (Demonstrations)},
  pages={54--59},
  year={2019}
}

---

Issues?

The Flair issue tracker is available here.