Bi-NER (bi_ner)

bi_ner is a arrangement of two encoder modules from BERT, it represents context and candidate separately using twin-structured encoders. You can train the module on custom data to behave accordingly. Let's take a deep dive into the training culture.

This tutorial shows you how to train a Bi-NER with a custom training loop to extracts entities from contexts. In this you use jaseci(jac) and python.

1. Import Bi-NER (bi_ner) module in jac
2. Train the model
3. Use the trained model to make predictions

Walk through

1. Import Bi-NER(bi_ner) module in jac

1. Open terminal and run jaseci by cmd

   jsctl -m

2. Load bi_ner module in jac by cmd

   actions load module jaseci_ai_kit.bi_ner

**2. Train and Predict **

For this tutorial, we are going to train and Predict the bi-ner for Named Entity Recognition.

• Creating Jac Program (train and infer bi_ner)

  1. Create a file by name bi_ner.jac

  2. Create train walker

     walker train{
         # import train ability
         can bi_ner.train;
         # the parameters required for training
         has train_file = "train_bi.json";
         has from_scratch = true;
         has num_train_epochs = 20;
         # load the file
         train_data = file.load_json(train_file);
         # Train the model
         report bi_ner.train(
             dataset=train_data,
             from_scratch=from_scratch,
             training_parameters={
                 "num_train_epochs": num_train_epochs.int
                 }
             );
         }
     

  3. Declaring Predict walker for extracting entities

     #
     walker predict{
         # passing input data for prediction
         has test_data_file = "test_dataset.json";
         test_data = file.load_json(test_data_file);
         report bi_ner.infer(
                         contexts=test_data,
                         );
     }
     

     Parameter details

     • train: will be used to train the Bi-NER on custom data
       • Input:
         • dataset (Dict): dictionary of candidates and suportting contexts for each candidate
         • from_scratch (bool): if set to true train the model from scratch otherwise trains incrementally
         • training_parameters (Dict): dictionary of training parameters
       • Returns: text when model training is completed
     • infer: will be used to predits the most suitable candidate for a provided context, takes text or embedding
       • Input:
         • contexts (string or list of strings): context which needs to be classified
       • Return: a dictionary of entities type and value with start and end index

  Steps for running bi_ner.jac programm

  4. Build bi_ner.jac by run cmd

     jac build bi_ner.jac

  5. Activate sentinal by run cmd

     sentinel set -snt active:sentinel -mode ir bi_ner.jir

  6. Calling walker train with default parameter for training bi_ner module by cmd

     walker run train
     

  7. Calling walker predict with default parameter to get evalution

     walker run predict
     

  Sample Input Data for Training

  {
      "text": [
          "IL-2 gene expression and NF-kappa B activation through CD28 requires reactive oxygen production by 5-lipoxygenase.",
          "Activation of the CD28 surface receptor provides a major costimulatory signal for T cell activation resulting in enhanced production of interleukin-2 (IL-2) and cell proliferation."
      ],
      "annotations": [
          [
              {
                  "start_index": 0,
                  "end_index": 9,
                  "entity_type": "G#DNA"
              },
              {
                  "start_index": 25,
                  "end_index": 35,
                  "entity_type": "G#protein"
              },
              {
                  "start_index": 55,
                  "end_index": 59,
                  "entity_type": "G#protein"
              },
              {
                  "start_index": 99,
                  "end_index": 113,
                  "entity_type": "G#protein"
              }
          ],
          [
              {
                  "start_index": 18,
                  "end_index": 22,
                  "entity_type": "G#protein"
              },
              {
                  "start_index": 18,
                  "end_index": 39,
                  "entity_type": "G#protein"
              },
              {
                  "start_index": 136,
                  "end_index": 149,
                  "entity_type": "G#protein"
              },
              {
                  "start_index": 151,
                  "end_index": 155,
                  "entity_type": "G#protein"
              }
          ]
      ]
  }