bwbayu/Flair_POS_FastText_id

English NER in Flair (default model)

This is the POS model for Indonesian that ships with Flair. The architecture of this model uses FastText.

F-score (micro) = 0.9345
F-score (macro) = 0.8735
Accuracy = 0.9345

Predicts 19 tags:

Tag	Meaning
NOUN	Noun (person, place, thing, or idea)
PROPN	Proper noun (specific name)
PUNCT	Punctuation (marks like commas, periods, etc.)
VERB	Verb (action or state)
ADP	Adposition (prepositions or postpositions)
PRON	Pronoun (substitute for a noun)
ADJ	Adjective (describes a noun)
NUM	Numeral (number or quantity)
DET	Determiner (a word that modifies a noun)
CCONJ	Coordinating conjunction (joins clauses or words)
ADV	Adverb (modifies a verb, adjective, or another adverb)
AUX	Auxiliary verb (helps the main verb)
SCONJ	Subordinating conjunction (introduces subordinate clauses)
PART	Particle (small word that doesn’t change in form, e.g., "not")
SYM	Symbol (mathematical or other special symbols)
X	Other (words that don't fit standard POS categories)
INTJ	Interjection (expresses strong emotion or reaction)

Demo: How to use in Flair

Requires: Flair (pip install flair).

You also need to download the model file locally to use it.

You can find training or fine-tuning code here : https://github.com/bwbayu/product_name_clustering/blob/main/additional/train_pos_flair.ipynb

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

tagger = SequenceTagger.load("model")
text = "aku pergi ke pasar"
sentence = Sentence(text)
tagger.predict(sentence)
for token in sentence:
    print(f"{token.text} ({token.get_label('upos').value})")

This yields the following output:

aku (PRON)
pergi (VERB)
ke (ADP)
pasar (NOUN)

Cite

Please cite the following paper when using this model.

@inproceedings{akbik2018coling,
  title={Contextual String Embeddings for Sequence Labeling},
  author={Akbik, Alan and Blythe, Duncan and Vollgraf, Roland},
  booktitle = {{COLING} 2018, 27th International Conference on Computational Linguistics},
  pages     = {1638--1649},
  year      = {2018}
}

Issues?

The Flair issue tracker is available here.