Arabic Lemmatizer

Description

This model converts words to their basic form. For example, it can convert past and present tense of a word, singular and plural words in a single form, which enables the downstream model to treat both words similarly instead of different words. The lemmatizer takes into consideration the context surrounding a word to determine which root is correct when the word form alone is ambiguous.

Open in Colab Download

How to use

Use as part of a pipeline after tokenisation.

...
lemmatizer = LemmatizerModel.pretrained("lemma", "ar") \
        .setInputCols(["token"]) \
        .setOutputCol("lemma")
nlp_pipeline = Pipeline(stages=[document_assembler, tokenizer, lemmatizer])
light_pipeline = LightPipeline(nlp_pipeline.fit(spark.createDataFrame([['']]).toDF("text")))
results = light_pipeline.fullAnnotate(['يرفض	رفضت	يرفضون	أرفض	نرفض	رفض	رفضوا	ترفض	رفضتا'])
...
val lemmatizer = LemmatizerModel.pretrained("lemma", "ar")
        .setInputCols(Array("token"))
        .setOutputCol("lemma")
val pipeline = new Pipeline().setStages(Array(document_assembler, tokenizer, lemmatizer))
val result = pipeline.fit(Seq.empty["يرفض	رفضت	يرفضون	أرفض	نرفض	رفض	رفضوا	ترفض	رفضتا"].toDS.toDF("text")).transform(data)
import nlu

text = ["""يرفض	رفضت	يرفضون	أرفض	نرفض	رفض	رفضوا	ترفض	رفضتا"""]
lemma_df = nlu.load('ar.lemma').predict(text, output_level='document')
lemma_df.lemma.values[0]

Results

{'lemma': [Annotation(token, 0, 3, رَفَض, {'sentence': '0'}),
   Annotation(token, 5, 8, رَفَض, {'sentence': '0'}),
   Annotation(token, 10, 15, رَفَض, {'sentence': '0'}),
   Annotation(token, 17, 20, رَفَض, {'sentence': '0'}),
   Annotation(token, 22, 25, نرفض, {'sentence': '0'}),
   Annotation(token, 27, 29, رَفض, {'sentence': '0'}),
   Annotation(token, 31, 35, رَفَض, {'sentence': '0'}),
   Annotation(token, 37, 40, رَفَض, {'sentence': '0'}),
   Annotation(token, 42, 46, رَفَض, {'sentence': '0'})]}

Model Information

Model Name: lemma
Compatibility: Spark NLP 2.7.0+
Edition: Official
Input Labels: [token]
Output Labels: [lemma]
Language: ar

Data Source

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