Named Entity Recognition for Korean (GloVe 840B 300d)

Description

This model annotates named entities in a text, that can be used to find features such as names of people, places, and organizations in the BIO format. The model does not read words directly but instead reads word embeddings, which represent words as points such that more semantically similar words are closer together.

This model uses the pre-trained glove_840B_300 embeddings model from WordEmbeddings annotator as an input, so be sure to use the same embeddings in the pipeline.

Predicted Entities

DT (date), LC (location), OG (organization), PS (person), TI (time), and O (other)

Download

How to use


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

sentence_detector = SentenceDetector()\
        .setInputCols(["document"])\
        .setOutputCol("sentence")

word_segmenter = WordSegmenterModel.pretrained("wordseg_kaist_ud", "ko")\
        .setInputCols(["sentence"])\
        .setOutputCol("token")

embeddings = WordEmbeddingsModel.pretrained("glove_840B_300", "xx")\
          .setInputCols("document", "token") \
          .setOutputCol("embeddings")


ner = NerDLModel.pretrained("ner_kmou_glove_840B_300d", "ko") \
        .setInputCols(["document", "token", "embeddings"]) \
        .setOutputCol("ner")

pipeline = Pipeline(stages=[
        document_assembler,
        sentence_detector,
        word_segmenter,
        embeddings,
        ner])

example = spark.createDataFrame(pd.DataFrame({'text': ["라이프니츠 의 주도 로 베를린 에 세우 어 지 ㄴ 베를린 과학아카데미"]}))

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

val sentence_detector = SentenceDetector()
        .setInputCols(["document"])
        .setOutputCol("sentence")

val word_segmenter = WordSegmenterModel.pretrained("wordseg_kaist_ud", "ko")
        .setInputCols(["sentence"])
        .setOutputCol("token")

val embeddings = WordEmbeddingsModel.pretrained("glove_840B_300", "xx")
          .setInputCols("document", "token")
          .setOutputCol("embeddings")


val ner = NerDLModel.pretrained("ner_kmou_glove_840B_300d", "ko")
        .setInputCols(["document", "token", "embeddings"])
        .setOutputCol("ner")

val pipeline = new Pipeline().setStages(Array(document_assembler, sentence_detector, word_segmenter, embeddings, ner))

val result = pipeline.fit(Seq.empty["라이프니츠  주도  베를린  세우    베를린 과학아카데미"].toDS.toDF("text")).transform(data)

Results

+------------+----+
|token       |ner |
+------------+----+
|라이프니츠  |B-PS|
|의          |O   |
|주도        |O   |
|로          |O   |
|베를린      |O   |
|에          |O   |
|세우        |O   |
|어          |O   |
|지          |O   |
|ㄴ          |O   |
|베를린      |B-OG|
|과학아카데미|I-OG|
+------------+----+

Model Information

Model Name: ner_kmou_glove_840B_300d
Type: ner
Compatibility: Spark NLP 2.7.0+
License: Open Source
Edition: Official
Input Labels: [sentence, token, embeddings]
Output Labels: [ner]
Language: ko

Data Source

The model was trained in the Korea Maritime and Ocean University NLP data set.

Benchmarking

|    ner_tag   | precision | recall | f1-score | support |
|:------------:|:---------:|:------:|:--------:|:-------:|
|     B-DT     |    0.95   |  0.29  |   0.44   |   132   |
|     B-LC     |    0.00   |  0.00  |   0.00   |   166   |
|     B-OG     |    1.00   |  0.06  |   0.11   |   149   |
|     B-PS     |    0.86   |  0.13  |   0.23   |   287   |
|     B-TI     |    0.50   |  0.05  |   0.09   |    20   |
|     I-DT     |    0.94   |  0.36  |   0.52   |   164   |
|     I-LC     |    0.00   |  0.00  |   0.00   |    4    |
|     I-OG     |    1.00   |  0.08  |   0.15   |    25   |
|     I-PS     |    1.00   |  0.08  |   0.15   |    12   |
|     I-TI     |    0.50   |  0.10  |   0.17   |    10   |
|       O      |    0.94   |  1.00  |   0.97   |  12830  |
|   accuracy   |    0.94   |  13799 |          |         |
|   macro avg  |    0.70   |  0.20  |   0.26   |  13799  |
| weighted avg |    0.93   |  0.94  |   0.92   |  13799  |