flair-uk-ner
Model description
flair-uk-pos is a Flair model that is ready to use for part-of-speech (upos) tagging. It is based on flair embeddings, that I've trained for Ukrainian language (available here and here) and has superior performance and a very small size (just 72mb!).
Results:
- F-score (micro) 0.9793
- F-score (macro) 0.9275
- Accuracy 0.9793
By class: | precision | recall | f1-score | support |
---|---|---|---|---|
NOUN | 0.9857 | 0.9851 | 0.9854 | 4549 |
PUNCT | 0.9984 | 1.0000 | 0.9992 | 3097 |
ADJ | 0.9772 | 0.9852 | 0.9812 | 1959 |
ADP | 0.9956 | 0.9968 | 0.9962 | 1584 |
VERB | 0.9891 | 0.9910 | 0.9900 | 1552 |
ADV | 0.9630 | 0.9118 | 0.9367 | 714 |
CCONJ | 0.9685 | 0.9746 | 0.9715 | 630 |
PROPN | 0.9279 | 0.9472 | 0.9375 | 625 |
DET | 0.9729 | 0.9698 | 0.9713 | 629 |
PRON | 0.9706 | 0.9631 | 0.9669 | 515 |
PART | 0.9235 | 0.8693 | 0.8956 | 375 |
NUM | 0.9722 | 0.9804 | 0.9763 | 357 |
SCONJ | 0.8768 | 0.9577 | 0.9154 | 260 |
AUX | 0.8906 | 0.9500 | 0.9194 | 120 |
X | 0.9833 | 0.9593 | 0.9712 | 123 |
SYM | 1.0000 | 0.7059 | 0.8276 | 17 |
INTJ | 0.5556 | 0.5000 | 0.5263 | 10 |
accuracy | 0.9793 | 17116 | ||
macro avg | 0.9383 | 0.9204 | 0.9275 | 17116 |
weighted avg | 0.9794 | 0.9793 | 0.9792 | 17116 |
The model was fine-tuned on the Ukrainian (UD) corpus, released by the non-profit organization Institute for Ukrainian. Training code is also available here.
Usage demo
from flair.data import Sentence
from flair.models import SequenceTagger
from pprint import pprint
tagger = SequenceTagger.load("dchaplinsky/flair-uk-pos")
sentence = Sentence("Я люблю Україну. Моє імʼя Марія Шевченко, я навчаюся в Київській політехніці.")
tagger.predict(sentence)
print(sentence)
print('---')
print('The following POS tags are found:')
pos_items = []
for label in sentence.get_labels():
all_labels = []
keys = label.data_point.annotation_layers.keys()
for key in keys:
all_labels.extend(
[
{'label': label.value, 'score': round(label.score, 4)}
for label in label.data_point.get_labels(key)
if label.data_point == label
]
)
pos_items.append({
'text': label.data_point.text,
'all_labels': all_labels,
})
pprint(pos_items)
# Result:
"""
Sentence: "Я люблю Україну . Моє імʼя Марія Шевченко , я навчаюся в Київській політехніці ." → ["Я"/PRON, "люблю"/VERB, "Україну"/PROPN, "."/PUNCT, "Моє"/DET, "імʼя"/NOUN, "Марія"/PROPN, "Шевченко"/PROPN, ","/PUNCT, "я"/PRON, "навчаюся"/VERB, "в"/ADP, "Київській"/ADJ, "політехніці"/NOUN, "."/PUNCT]
---
The following POS tags are found:
[{'all_labels': [{'label': 'PRON', 'score': 1.0}], 'text': 'Я'},
{'all_labels': [{'label': 'VERB', 'score': 1.0}], 'text': 'люблю'},
{'all_labels': [{'label': 'PROPN', 'score': 1.0}], 'text': 'Україну'},
{'all_labels': [{'label': 'PUNCT', 'score': 1.0}], 'text': '.'},
{'all_labels': [{'label': 'DET', 'score': 0.9999}], 'text': 'Моє'},
{'all_labels': [{'label': 'NOUN', 'score': 1.0}], 'text': 'імʼя'},
{'all_labels': [{'label': 'PROPN', 'score': 1.0}], 'text': 'Марія'},
{'all_labels': [{'label': 'PROPN', 'score': 1.0}], 'text': 'Шевченко'},
{'all_labels': [{'label': 'PUNCT', 'score': 1.0}], 'text': ','},
{'all_labels': [{'label': 'PRON', 'score': 1.0}], 'text': 'я'},
{'all_labels': [{'label': 'VERB', 'score': 1.0}], 'text': 'навчаюся'},
{'all_labels': [{'label': 'ADP', 'score': 1.0}], 'text': 'в'},
{'all_labels': [{'label': 'ADJ', 'score': 1.0}], 'text': 'Київській'},
{'all_labels': [{'label': 'NOUN', 'score': 1.0}], 'text': 'політехніці'},
{'all_labels': [{'label': 'PUNCT', 'score': 1.0}], 'text': '.'}]
"""
Copyright: Dmytro Chaplynskyi, lang-uk project, 2022
- Downloads last month
- 5
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Evaluation results
- POS F Scoreself-reported0.979