/NLP_NER_BERT_BILSTM_CRF_Named_Entity_Recognition

NLP Named Entity Recognition dalam bidang Biomedis, mendeteksi teks dan membuat klasifikasi apakah teks tersebut mempunyai entitas plant atau disease, memberi label pada teks, menguji hubungan entitas plant dan disease, menilai kecocokan antara kedua entitas, membandingkan hasil uji dengan menggunakan models BERT-BILSTM-CRF

Primary LanguageJupyter Notebook

NLP

Named Entity Recognition (NER) - BERT - BILSTM -CRF

Slamet Riyanto S.Kom., M.M.S.I.

Dimas Dwi Putra

Architecture

Dataset
View Dataset .json

B-I-O

{0: 'O', 1: 'B', 2: 'I'}

Labels

{'UNK': 0, 'plant': 1, 'disease': 2} {0: 'UNK', 1: 'plant', 2: 'disease'}

Json

[
    {
      'id' : ...,
      'labels' : ['Truncate', 'Type Entity', 'Start Entity', 'End Entity', 'Entity Names'],
      'text': ...
    }
]

Example

[
    {
      'id': 0,
      'labels': [['T0', 'plant', 46, 55, 'digitalis'],
       ['T1', 'disease', 64, 75, 'arrhythmias']],
      'text': 'studies on magnesium s mechanism of action in digitalis induced arrhythmias'
    },
    ...
]

Eval
View Eval Report .xlsx

Fine Tuning Biobert-Plant-Disease Biobert-Plant-Disease Biobert-Plant-Disease Biobert-Plant-Disease
Model Bert Bert-CRF Bert-Bilstm Bert-Bilstm-CRF
Batch Size 2 2 2 2
Epoch 10 10 10 10
Iterasi 393 393 393 393
Step 3.930 3.930 3.930 3.930
Learning Rate 0,00003 0,00003 0,00003 0,00003
Dropout 0,1 0,1 0,1 0,1
Entitas (Plant) (Disease) (Plant) (Disease) (Plant) (Disease) (Plant) (Disease)
Precision (0,86) (0,66) (0,79) (0,64) (0,87) (0,68) (0,82) (0,62)
Recall (0,64) (0,43) (0,64) (0,41) (0,64) (0,42) (0,64) (0,44)
F-1 Score (0,73) (0,52) (0,71) (0,5) (0,74) (0,51) (0,72) (0,51)
Average/Total
Precision 0,74 0,71 0,76 0,7
Recall 0,51 0,5 0,51 0,52
F-1 Score 0,61 0,58 0,61 0,6
Eksekusi 0:22:35 1:01:40 0:25:28 1:04:07
Device Cuda Tesla T4 Cuda Tesla T4 Cuda Tesla T4 Cuda Tesla T4

Predict

effects of korean red ginseng extracts on neural tube defects and impairment of social interaction induced by prenatal exposure to valproic
{'plant': [['ginseng', 22]], 'disease': [['neural', 42], ['tube', 49], ['defects', 54]]}

Output
View Model Directory

Save model output as Pytorch .pt

Other Content

Websites Prediction

1. Django Websites Prediction For NER dan RE

Named Entity Recognition (NER)

1. NER Dataset Biomedical Plant-Disease Corpus

2. NER CRF Named Entity Recognition

3. NER BiLSTM Named Entity Recognition

4. NER BERT Named Entity Recognition

5. NER BiLSTM CRF Named Entity Recognition

6. NER BERT BiLSTM CRF Named Entity Recognition

Relation Extraction (RE)

1. RE Dataset Biomedical Plant-Disease Corpus

2. RE BERT Relation Extraction Biomedical

3. RE BiLSTM CRF Relation Extraction Biomedical