Part of Speech for Urdu

Description

This model annotates the part of speech of tokens in a text. The parts of speech annotated include PRON (pronoun), CCONJ (coordinating conjunction), and 15 others. The part of speech model is useful for extracting the grammatical structure of a piece of text automatically.

Open in Colab Download

How to use

...
pos = PerceptronModel.pretrained("pos_ud_udtb", "ur") \
    .setInputCols(["document", "token"]) \
    .setOutputCol("pos")
nlp_pipeline = Pipeline(stages=[document_assembler, sentence_detector, tokenizer, pos])
light_pipeline = LightPipeline(nlp_pipeline.fit(spark.createDataFrame([['']]).toDF("text")))
results = light_pipeline.fullAnnotate(["شمال کا بادشاہ ہونے کے علاوہ ، جان سن ایک انگریزی معالج ہے۔"])

...
val pos = PerceptronModel.pretrained("pos_ud_udtb", "ur")
    .setInputCols(Array("document", "token"))
    .setOutputCol("pos")
val pipeline = new Pipeline().setStages(Array(document_assembler, sentence_detector, tokenizer, pos))
val result = pipeline.fit(Seq.empty["شمال کا بادشاہ ہونے کے علاوہ ، جان سن ایک انگریزی معالج ہے۔"].toDS.toDF("text")).transform(data)

Results

{'pos': [Annotation(pos, 0, 3, NOUN, {'word': 'شمال'}),
   Annotation(pos, 5, 6, ADP, {'word': 'کا'}),
   Annotation(pos, 8, 13, NOUN, {'word': 'بادشاہ'}),
   Annotation(pos, 15, 18, VERB, {'word': 'ہونے'}),
   Annotation(pos, 20, 21, ADP, {'word': 'کے'}),
   Annotation(pos, 23, 27, ADP, {'word': 'علاوہ'}),
   Annotation(pos, 29, 29, PUNCT, {'word': '،'}),
   Annotation(pos, 31, 33, PROPN, {'word': 'جان'}),
   Annotation(pos, 35, 36, PROPN, {'word': 'سن'}),
   Annotation(pos, 38, 40, NUM, {'word': 'ایک'}),
   Annotation(pos, 42, 48, PROPN, {'word': 'انگریزی'}),
   Annotation(pos, 50, 54, ADJ, {'word': 'معالج'}),
   Annotation(pos, 56, 58, PUNCT, {'word': 'ہے۔'})]}

Model Information

Model Name: pos_ud_udtb
Compatibility: Spark NLP 2.7.0+
Edition: Official
Input Labels: [tags, document]
Output Labels: [pos]
Language: ur

Data Source

The model is trained on data obtained from https://universaldependencies.org

Benchmarking

|    |              | precision   | recall   |   f1-score |   support |
|---:|:-------------|:------------|:---------|-----------:|----------:|
|  0 | ADJ          | 0.84        | 0.85     |       0.84 |      1117 |
|  1 | ADP          | 0.98        | 0.99     |       0.98 |      3122 |
|  2 | ADV          | 0.83        | 0.65     |       0.73 |       125 |
|  3 | AUX          | 0.97        | 0.96     |       0.96 |       937 |
|  4 | CCONJ        | 0.96        | 1.00     |       0.98 |       338 |
|  5 | DET          | 0.87        | 0.82     |       0.84 |       237 |
|  6 | NOUN         | 0.89        | 0.92     |       0.9  |      3690 |
|  7 | NUM          | 0.97        | 0.95     |       0.96 |       267 |
|  8 | PART         | 0.96        | 0.88     |       0.91 |       337 |
|  9 | PRON         | 0.96        | 0.94     |       0.95 |       499 |
| 10 | PROPN        | 0.88        | 0.85     |       0.86 |      1975 |
| 11 | PUNCT        | 1.00        | 1.00     |       1    |       682 |
| 12 | SCONJ        | 0.97        | 0.99     |       0.98 |       248 |
| 13 | VERB         | 0.95        | 0.95     |       0.95 |      1232 |
| 14 | accuracy     |             |          |       0.93 |     14806 |
| 15 | macro avg    | 0.93        | 0.91     |       0.92 |     14806 |
| 16 | weighted avg | 0.93        | 0.93     |       0.93 |     14806 |
​