සිංSRL-Direct Annotator

1. First satisfy the following requirements

   python 3.6
tensorflow 1.4.0
gensim 3.4.0
flask-cors

2. Training data should be in following directories

   The predicate identifier - /data/predIdData
   The srl tagger - /data/srlIdData

3. Fasttext word embedding model should be in /data directory

4. To train the model execute following script

   The predicate identifier - python train.py predId train
   The srl tagger - python train.py srlId train

5. If training was interrupted. It can be continued again using following command

   python train.py predId retrain {{modelname}}
python train.py srlId retrain {{modelname}}
   model name here should be the directory name that saves checkpoints for the model that should continue with training. They can be found at root/results/test/predIdData/model.weights
root/results/test/srlIdData/model.weights

6. Training parameters can be configured using parameters.json file which available in the project root directory

7. To get the predictions from each trained models -

   predicate identifier - python individualModelpredict.py predId {{modelname}}
   srl tagger - python individualModelpredict.py srlId {{modelname}}

8.To get the final prediction for a given Sinhala sentence

`python finalPredict.py {{predId model name}} {{srlId model name}}`  

9. logs are stored in /results directory

NOTE: ServerVersionDirectSinSRL directory consists the REST implementation of the project

When run in COLAB

execute following code snippet inside the browser console to remain connected the colab session

function ClickConnect(){
    console.log("Working"); 
    document.querySelector("colab-connect-button").click() 
}
setInterval(ClickConnect,60000)

View the previous research