/*
* LogWordPriorFileParser.java
*
* Copyright 2006 Michigan State University Board of Trustees
*
* Created on September 18, 2003, 5:38 PM
*/
package edu.msu.cme.rdp.classifier.io;
import edu.msu.cme.rdp.classifier.TrainingDataException;
import edu.msu.cme.rdp.classifier.utils.HierarchyVersion;
import java.io.*;
import java.util.StringTokenizer;
/**
* A parser to parse a read containing a list of the indices of words
* and the log value of word prior probabilities.
* @author wangqion
*/
public class LogWordPriorFileParser {
/** Reads a file and saves the log word priors in an array.
* Returns the version information for validation purpose.
* Input file format: integer follows by a tab and float each line.
* The first value indicates the index of a word, the second value
* indicates the log value of word prior probability.
*/
public static HierarchyVersion createLogWordPriorArr(Reader r, float[] arr, HierarchyVersion version) throws IOException, TrainingDataException {
BufferedReader reader = new BufferedReader(r);
String line = reader.readLine();
if (line != null) {
HierarchyVersion thisVersion = new HierarchyVersion(line);
int trainsetNo = thisVersion.getTrainsetNo();
if (thisVersion.getVersion() == null) {
throw new TrainingDataException("Error: There is no version information "
+ "in the wordPrior file");
}
if (version == null) {
version = thisVersion;
} else if (!version.getVersion().equals(thisVersion.getVersion()) || version.getTrainsetNo() != thisVersion.getTrainsetNo()) {
throw new TrainingDataException("Error: The version information in the wordPrior file is different from the version of the other training files.");
}
}
while ((line = reader.readLine()) != null) {
StringTokenizer st = new StringTokenizer(line, "\t");
if (st.countTokens() != 2) {
throw new TrainingDataException("\nError: " + line + " does not have exact two numbers");
}
try {
int wordIndex = Integer.parseInt(st.nextToken());
float logWordPrior = Float.parseFloat(st.nextToken());
arr[wordIndex] = logWordPrior;
} catch (NumberFormatException e) {
reader.close();
throw new TrainingDataException("\nError: "
+ "The value for wordIndex or word prior is not a number at line : " + line);
}
}
reader.close();
return version;
}
}