LDAAnalyzer.java

package de.vipra.cmd.lda;

import java.util.List;

import de.vipra.cmd.ex.LDAAnalyzerException;
import de.vipra.util.Config;
import de.vipra.util.Config.Key;
import de.vipra.util.Constants;
import de.vipra.util.ConvertStream;
import de.vipra.util.WordMap;
import de.vipra.util.model.TopicFull;
import de.vipra.util.model.TopicRef;

public abstract class LDAAnalyzer {

	private final String name;

	protected LDAAnalyzer(String name) {
		this.name = name;
	}

	public String getName() {
		return name;
	}

	public abstract void init(Config config, WordMap wordMap) throws LDAAnalyzerException;

	public abstract void analyze() throws LDAAnalyzerException;

	/**
	 * Returns a converting stream of topics, read from the topic definition
	 * file. Usually, a topic definition consists of a list of words, that are
	 * assigned to that topic with a certain likeliness.
	 * 
	 * @return topic definition stream
	 * @throws LDAAnalyzerException
	 */
	public abstract ConvertStream<TopicFull> getTopicDefinitions() throws LDAAnalyzerException;

	/**
	 * Returns a converting stream of lists of topic references. Normally, topic
	 * modeling outputs topics for each word of each document. These references
	 * are returned by this function.
	 * 
	 * @return stream of lists of topic references per document (ordered by
	 *         index)
	 * @throws LDAAnalyzerException
	 */
	public abstract ConvertStream<List<TopicRef>> getTopics() throws LDAAnalyzerException;

	public static LDAAnalyzer getAnalyzer(Config config, WordMap wordMap) throws LDAAnalyzerException {
		LDAAnalyzer analyzer = null;
		switch (Constants.Analyzer.fromString(config.getString(Key.ANALYZER))) {
			case JGIBB:
			case DEFAULT:
			default:
				analyzer = new JGibbLDAAnalyzer();
				break;
		}
		analyzer.init(config, wordMap);
		return analyzer;
	}

}