private void explore(LineReader grammar) {
int counter = 0;
boolean first_line = true;
while (grammar.hasNext()) {
String line = grammar.next().trim();
counter++;
String[] fields = line.split("\\s\\|{3}\\s");
if (fields.length < 4) {
logger.warning("Incomplete grammar line at line " + counter);
continue;
}
String lhs = fields[0];
String[] source = fields[1].split("\\s");
String[] target = fields[2].split("\\s");
String[] features = fields[3].split("\\s");
Vocabulary.id(lhs);
// Add symbols to vocabulary.
for (String source_word : source) {
if (FormatUtils.isNonterminal(source_word)) Vocabulary.id(FormatUtils.stripNt(source_word));
else Vocabulary.id(source_word);
}
for (String target_word : target) {
if (FormatUtils.isNonterminal(target_word)) Vocabulary.id(FormatUtils.stripNt(target_word));
else Vocabulary.id(target_word);
}
// Test features for labeling.
if (first_line && features.length != 0) {
if (!features[0].contains("=")) {
// We assume that if there is one unlabeled feature the entire grammar is unlabeled.
labeled = false;
}
this.types.setLabeled(labeled);
first_line = false;
}
// Add feature names to vocabulary and pass the value through the
// appropriate encoder.
for (int f = 0; f < features.length; ++f) {
if (labeled) {
String[] fe = features[f].split("=");
if (fe[0].equals("Alignment")) continue;
types.observe(Vocabulary.id(fe[0]), Float.parseFloat(fe[1]));
} else {
types.observe(f, Float.parseFloat(features[f]));
}
}
}
}
/**
* Executes the packing.
*
* @throws IOException
*/
public void pack() throws IOException {
logger.info("Beginning exploration pass.");
LineReader grammar_reader = null;
LineReader alignment_reader = null;
// Explore pass. Learn vocabulary and feature value histograms.
logger.info("Exploring: " + grammar);
grammar_reader = new LineReader(grammar);
explore(grammar_reader);
logger.info("Exploration pass complete. Freezing vocabulary and finalizing encoders.");
if (dump != null) {
PrintWriter dump_writer = new PrintWriter(dump);
dump_writer.println(types.toString());
dump_writer.close();
}
types.inferTypes(this.labeled);
logger.info("Type inference complete.");
logger.info("Finalizing encoding.");
logger.info("Writing encoding.");
types.write(output + File.separator + "encoding");
logger.info("Freezing vocab.");
Vocabulary.freeze();
logger.info("Writing vocab.");
Vocabulary.write(output + File.separator + "vocabulary");
// Read previously written encoder configuration to match up to changed
// vocabulary id's.
logger.info("Reading encoding.");
encoderConfig = new EncoderConfiguration();
encoderConfig.load(output + File.separator + "encoding");
logger.info("Beginning packing pass.");
Queue<PackingFileTuple> slices = new PriorityQueue<PackingFileTuple>();
// Actual binarization pass. Slice and pack source, target and data.
grammar_reader = new LineReader(grammar);
if (packAlignments) alignment_reader = new LineReader(alignments);
binarize(grammar_reader, alignment_reader, slices);
logger.info("Packing complete.");
logger.info("Packed grammar in: " + output);
logger.info("Done.");
}
public GrammarPacker(
String grammar_filename,
String config_filename,
String output_filename,
String alignments_filename,
String featuredump_filename)
throws IOException {
this.labeled = true;
this.grammar = grammar_filename;
this.output = output_filename;
this.dump = featuredump_filename;
// TODO: Always open encoder config? This is debatable.
this.types = new FeatureTypeAnalyzer(true);
this.alignments = alignments_filename;
packAlignments = (alignments != null);
if (!packAlignments) {
logger.info("No alignments file specified, skipping.");
} else if (!new File(alignments_filename).exists()) {
logger.severe("Alignments file does not exist: " + alignments);
System.exit(0);
}
if (config_filename != null) {
readConfig(config_filename);
types.readConfig(config_filename);
} else {
logger.info("No config specified. Attempting auto-detection of feature types.");
}
File working_dir = new File(output);
working_dir.mkdir();
if (!working_dir.exists()) {
logger.severe("Failed creating output directory.");
System.exit(0);
}
}
