Loading the model only one time!

[Norio Kosaka]

I am thinking to implement this brilliant library into my project.
But when it comes to the production level coding, one thing on my mind, which is loading model.

Could you tell us that how we just need to load the model only 1 time.

" One time" in a sense, once we load the model then we retain it while classifying the text.

For instance,

1. load the model
2. get text and classify it
3. retain the model on memory --> go back to 2

The thing I know is the function below.
// Load model from file
jft.loadModel("src/test/resources/models/supervised.model.bin");

If we succeed to do this, it will save our resource to classify a new message without loading the model to memory again.

Best,

[ashuezy]

Its quite easy::

import java.util.ArrayList;
import java.util.List;

import org.apache.log4j.Logger;

import com.github.jfasttext.JFastText;
import com.github.jfasttext.JFastText.ProbLabel;
import com.xyz.constants.SentimentConstants;

public class TaxonomyUtil
{
	private static final Logger logger = Logger.getLogger(TaxonomyUtil.class);
	private static JFastText jft = new JFastText();
	
	public static void init()
	{
		jft.loadModel(PathUtil.getPath(SentimentConstants.TAXONOMY_FILE_PATH));
		System.out.println("Taxonomy Model -done");
	}
	
	public static ArrayList<String> getLabel(String entireData)
	{
		ArrayList<String> list = new ArrayList<String>();
		String label="";
		try
		{
			List<ProbLabel> predictionList = jft.predictProba(entireData, 1);
			if(predictionList.size()>0)
			{
				label = predictionList.get(0).label.replaceAll("__label__", "");
				list.add(label);
			}
		}
		catch(Exception e)
		{
			logger.error(e);
			e.printStackTrace();
		}
		return list;
	}
	
	public static void main(String args[])
	{
		init();
		System.out.println(getLabel("how are you"));
	}
}

Just call init at server startup one time Then call the getLabel method for every classification.

[Norio Kosaka]

Thank you for your help!!