Welsh Lemmatizer

Description

This is a dictionary-based lemmatizer that assigns all forms and inflections of a word to a single root. This enables the pipeline to treat the past and present tense of a verb, for example, as the same word instead of two completely different words.

Live Demo Open in Colab Download

How to use

document_assembler = DocumentAssembler() \
    .setInputCol("text") \
    .setOutputCol("document")

tokenizer = Tokenizer()\
    .setInputCols(["document"]) \
    .setOutputCol("token")

lemmatizer = LemmatizerModel.pretrained("lemma", "cy") \
        .setInputCols(["token"]) \
        .setOutputCol("lemma")

pipeline = Pipeline(stages=[document_assembler, tokenizer, lemmatizer])

example = spark.createDataFrame([["Dywedir yn aml taw rygbi 'r undeb yw mabolgamp genedlaethol Cymru , er mae pêl-droed yn denu mwy o wylwyr i 'r maes ."]], ["text"])

results = pipeline.fit(example).transform(example)
val document_assembler = DocumentAssembler()
    .setInputCol("text")
    .setOutputCol("document")

val tokenizer = Tokenizer()
    .setInputCols(["document"])
    .setOutputCol("token")

val lemmatizer = LemmatizerModel.pretrained("lemma", "cy")
        .setInputCols(["token"])
        .setOutputCol("lemma")

val pipeline = new Pipeline().setStages(Array(document_assembler, tokenizer, lemmatizer))
val data = Seq("Dywedir yn aml taw rygbi "r undeb yw mabolgamp genedlaethol Cymru , er mae pêl-droed yn denu mwy o wylwyr i "r maes .").toDF("text")
val result = pipeline.fit(data).transform(data)

import nlu

text = ["Dywedir yn aml taw rygbi 'r undeb yw mabolgamp genedlaethol Cymru , er mae pêl-droed yn denu mwy o wylwyr i 'r maes ."]
lemma_df = nlu.load('cy.lemma').predict(text, output_level = "document")
lemma_df.lemma.values[0]

Results

+------------+
|       lemma|
+------------+
|     Dywedir|
|          yn|
|         aml|
|         taw|
|       rygbi|
|           '|
|           r|
|       undeb|
|         bod|
|   mabolgamp|
|cenedlaethol|
|       Cymru|
|           ,|
|          er|
|         bod|
|   pêl-droed|
|          yn|
|        denu|
|        mawr|
|           o|
+------------+
only showing top 20 rows

Model Information

Model Name: lemma
Compatibility: Spark NLP 3.0.0+
License: Open Source
Edition: Official
Input Labels: [token]
Output Labels: [lemma]
Language: cy

Data Source

The model was trained on the Universal Dependencies version 2.7.

Benchmarking

Precision=0.74, Recall=0.71, F1-score=0.72