NER TF 2.0 implementation for CoNLL-2003 NER dataset

A detailed blog for this implementation is available at

https://medium.com/analytics-vidhya/ner-tensorflow-2-2-0-9f10dcf5a0a

'data' folder has the dataset. Model output files will be written in the 'model_ouput; directory.

To train the model

python3 train.py --data data --output  model_output --overwrite True 

Predict on test Dataset

python3 predict.py --data data --model_dir model_output

Predict a single sentence - Assign the sentence to test_sentence variable.

python3 predict.py --data data --model_dir model_output --predsingle True

REST-API

NER model deployed as REST API

python app.py

API will be live at 0.0.0.0:8000 endpoint predict

cURL request

curl -X POST http://0.0.0.0:8000/predict -H 'Content-Type: application/json' -d '{ "text": "John is in New York" }'

Output

{
    "result": [
        {
            "word": "John",
            "tag": "B-PER",
            
        },
        {
            
            "word": "is",
            "tag": "O"
        },
        {
            
            "word": "in",
            "tag": "O"
        },
         {
            "word": "New",
            "tag": "B-LOC",
            
        },
        {
            "word": "York",
            "tag": "I-LOC",
            
        }
        
    ]
}

Visualisations 
```bash
tensorboard --logdir=model_output/logs/train --port=6006 --bind_all
tensorboard --logdir=model_output/logs/valid --port=6006 --bind_all