Part of Speech for Bhojpuri (pos_ud_bhtb)

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 14 others. The part of speech model is useful for extracting the grammatical structure of a piece of text automatically.

Predicted Entities

ADJ, ADP, ADV, AUX, CCONJ, DET, INTJ, NOUN, NUM, PART, PRON, PROPN, PUNCT, SCONJ, VERB, and X

Live Demo Open in Colab Download Copy S3 URI

How to use

Use as part of an nlp pipeline after tokenization.

document_assembler = DocumentAssembler() \
    .setInputCol("text") \
    .setOutputCol("document")
    
sentence_detector = SentenceDetector()\
    .setInputCols(["document"])\
    .setOutputCol("sentence")
    
tokenizer = Tokenizer()\
    .setInputCols(["sentence"])\
    .setOutputCol("token")
        
pos = PerceptronModel.pretrained("pos_ud_bhtb", "bh") \
    .setInputCols(["document", "token"]) \
    .setOutputCol("pos")
pipeline = Pipeline(stages=[
        document_assembler,
        sentence_detector,
        tokenizer,
        pos
    ])

example = spark.createDataFrame([['ओहु लोग के मालूम बा कि श्लील होखते भोजपुरी के नींव हिल जाई ।']], ["text"])
result = pipeline.fit(example).transform(example)
val document_assembler = DocumentAssembler()
        .setInputCol("text")
        .setOutputCol("document")
        
val sentence_detector = SentenceDetector()
        .setInputCols("document")
        .setOutputCol("sentence")
        
val tokenizer = Tokenizer()
    .setInputCols("sentence")
    .setOutputCol("token")
        
val pos = PerceptronModel.pretrained("pos_ud_bhtb", "bh")
    .setInputCols(Array("document", "token"))
    .setOutputCol("pos")

val pipeline = new Pipeline().setStages(Array(document_assembler, sentence_detector, tokenizer, pos))
val data = Seq("ओहु लोग के मालूम बा कि श्लील होखते भोजपुरी के नींव हिल जाई ।").toDF("text")
val result = pipeline.fit(data).transform(data)
import nlu

text = ["ओहु लोग के मालूम बा कि श्लील होखते भोजपुरी के नींव हिल जाई ।"]
pos_df = nlu.load('bh.pos').predict(text)
pos_df

Results

+------------------------------------------------------------+----------------------------------------------------------------------------------+
|text                                                        |result                                                                            |
+------------------------------------------------------------+----------------------------------------------------------------------------------+
|ओहु लोग के मालूम बा कि श्लील होखते भोजपुरी के नींव हिल जाई ।|[DET, NOUN, ADP, NOUN, VERB, SCONJ, ADJ, VERB, PROPN, ADP, NOUN, VERB, AUX, PUNCT]|
+------------------------------------------------------------+----------------------------------------------------------------------------------+

Model Information

Model Name: pos_ud_bhtb
Compatibility: Spark NLP 2.7.0+
Edition: Official
Input Labels: [sentence, token]
Output Labels: [pos]
Language: bh

Data Source

The model was trained on the Universal Dependencies version 2.7.

Reference:

  • Ojha, A. K., & Zeman, D. (2020). Universal Dependency Treebanks for Low-Resource Indian Languages: The Case of Bhojpuri. Proceedings of the WILDRE5{–} 5th Workshop on Indian Language Data: Resources and Evaluation.

Benchmarking

|     pos     | precision | recall | f1-score | support |
|--------------|-----------|--------|----------|---------|
| ADJ          | 0.92      | 0.92   | 0.92     | 250     |
| ADP          | 0.95      | 0.96   | 0.96     | 989     |
| ADV          | 0.85      | 0.88   | 0.86     | 32      |
| AUX          | 0.93      | 0.95   | 0.94     | 355     |
| CCONJ        | 0.95      | 0.95   | 0.95     | 151     |
| DET          | 0.96      | 0.95   | 0.95     | 353     |
| INTJ         | 1.00      | 1.00   | 1.00     | 5       |
| NOUN         | 0.95      | 0.96   | 0.96     | 1854    |
| NUM          | 0.97      | 0.98   | 0.97     | 149     |
| PART         | 0.94      | 0.93   | 0.93     | 192     |
| PRON         | 0.95      | 0.94   | 0.95     | 335     |
| PROPN        | 0.94      | 0.94   | 0.94     | 419     |
| PUNCT        | 0.97      | 0.96   | 0.96     | 695     |
| SCONJ        | 1.00      | 0.96   | 0.98     | 118     |
| VERB         | 0.95      | 0.93   | 0.94     | 767     |
| X            | 0.50      | 1.00   | 0.67     | 1       |
| accuracy     |           |        | 0.95     | 6665    |
| macro avg    | 0.92      | 0.95   | 0.93     | 6665    |
| weighted avg | 0.95      | 0.95   | 0.95     | 6665    |