From 879536d8e7bec518af716d4ee425fcd0eda36937 Mon Sep 17 00:00:00 2001
From: "trevor.cohn" <trevor.cohn@ec762483-ff6d-05da-a07a-a48fb63a330f>
Date: Sat, 17 Jul 2010 03:13:09 +0000
Subject: Extra options

git-svn-id: https://ws10smt.googlecode.com/svn/trunk@305 ec762483-ff6d-05da-a07a-a48fb63a330f
---
 .../prjava/src/phrase/Corpus.java                  | 82 +++++++++++++++++++---
 .../prjava/src/phrase/PhraseCluster.java           |  2 +-
 .../prjava/src/phrase/PhraseCorpus.java            |  2 +-
 .../prjava/src/phrase/Trainer.java                 | 10 ++-
 4 files changed, 85 insertions(+), 11 deletions(-)

(limited to 'gi/posterior-regularisation/prjava')

diff --git a/gi/posterior-regularisation/prjava/src/phrase/Corpus.java b/gi/posterior-regularisation/prjava/src/phrase/Corpus.java
index f2c6b132..2afc18dc 100644
--- a/gi/posterior-regularisation/prjava/src/phrase/Corpus.java
+++ b/gi/posterior-regularisation/prjava/src/phrase/Corpus.java
@@ -18,7 +18,9 @@ public class Corpus
 	public int splitSentinel;
 	public int phraseSentinel;
 	public int rareSentinel;
-	private boolean[] wordIsRare;
+	private boolean[] rareWords;
+	private boolean[] rarePhrases;
+	private boolean[] rareContexts;
 
 	public Corpus()
 	{
@@ -62,7 +64,8 @@ public class Corpus
 		public TIntArrayList getRawContext()
 		{
 			return Corpus.this.getRawContext(contextId);
-		}		public String getContextString(boolean insertPhraseSentinel)
+		}		
+		public String getContextString(boolean insertPhraseSentinel)
 		{
 			return Corpus.this.getContextString(contextId, insertPhraseSentinel);
 		}
@@ -143,12 +146,12 @@ public class Corpus
 	public TIntArrayList getPhrase(int phraseId)
 	{
 		TIntArrayList phrase = phraseLexicon.lookup(phraseId);
-		if (wordIsRare != null)
+		if (rareWords != null)
 		{
 			boolean first = true;
 			for (int i = 0; i < phrase.size(); ++i)
 			{
-				if (wordIsRare[phrase.get(i)])
+				if (rareWords[phrase.get(i)])
 				{
 					if (first)
 					{
@@ -182,12 +185,12 @@ public class Corpus
 	public TIntArrayList getContext(int contextId)
 	{
 		TIntArrayList context = contextLexicon.lookup(contextId);
-		if (wordIsRare != null)
+		if (rareWords != null)
 		{
 			boolean first = true;
 			for (int i = 0; i < context.size(); ++i)
 			{
-				if (wordIsRare[context.get(i)])
+				if (rareWords[context.get(i)])
 				{
 					if (first)
 					{
@@ -319,8 +322,71 @@ public class Corpus
 				counts[context.get(i)] += e.getCount();
 		}
 
-		wordIsRare = new boolean[wordLexicon.size()];
+		int count = 0;
+		rareWords = new boolean[wordLexicon.size()];
 		for (int i = 0; i < wordLexicon.size(); ++i)
-			wordIsRare[i] = counts[i] < wordThreshold;
+		{
+			rareWords[i] = counts[i] < wordThreshold;
+			if (rareWords[i])
+				count++;
+		}
+		System.err.println("There are " + count + " rare words");
+	}
+	
+	public void applyPhraseThreshold(int threshold) 
+	{
+		rarePhrases = new boolean[phraseLexicon.size()];
+
+		int n = 0;
+		for (int i = 0; i < phraseLexicon.size(); ++i)
+		{
+			List<Edge> contexts = phraseToContext.get(i);
+			int count = 0;
+			for (Edge edge: contexts)
+			{
+				count += edge.getCount();
+				if (count >= threshold)
+					break;
+			}
+			
+			if (count < threshold)
+			{
+				rarePhrases[i] = true;
+				n++;
+			}
+		}
+		System.err.println("There are " + n + " rare phrases");
+	}
+	
+	public void applyContextThreshold(int threshold) 
+	{
+		rareContexts = new boolean[contextLexicon.size()];
+
+		int n = 0;
+		for (int i = 0; i < contextLexicon.size(); ++i)
+		{
+			List<Edge> phrases = contextToPhrase.get(i);
+			int count = 0;
+			for (Edge edge: phrases)
+			{
+				count += edge.getCount();
+				if (count >= threshold)
+					break;
+			}
+			
+			if (count < threshold)
+			{
+				rareContexts[i] = true;
+				n++;
+			}
+		}
+		System.err.println("There are " + n + " rare contexts");
+	}
+	
+	boolean isRare(Edge edge)
+	{
+		if (rarePhrases != null && rarePhrases[edge.getPhraseId()] == true) return true;
+		if (rareContexts != null && rareContexts[edge.getContextId()] == true) return true;
+		return false;
 	}
 }
\ No newline at end of file
diff --git a/gi/posterior-regularisation/prjava/src/phrase/PhraseCluster.java b/gi/posterior-regularisation/prjava/src/phrase/PhraseCluster.java
index feab5eda..9ee766d4 100644
--- a/gi/posterior-regularisation/prjava/src/phrase/PhraseCluster.java
+++ b/gi/posterior-regularisation/prjava/src/phrase/PhraseCluster.java
@@ -103,7 +103,7 @@ public class PhraseCluster {
 			for (int ctx=0; ctx<contexts.size(); ctx++)
 			{
 				Edge edge = contexts.get(ctx);
-				if (edge.getCount() < edge_threshold) 
+				if (edge.getCount() < edge_threshold || c.isRare(edge)) 
 					continue;
 				
 				double p[]=posterior(edge);
diff --git a/gi/posterior-regularisation/prjava/src/phrase/PhraseCorpus.java b/gi/posterior-regularisation/prjava/src/phrase/PhraseCorpus.java
index 903e47c8..0cf31c1c 100644
--- a/gi/posterior-regularisation/prjava/src/phrase/PhraseCorpus.java
+++ b/gi/posterior-regularisation/prjava/src/phrase/PhraseCorpus.java
@@ -74,7 +74,7 @@ public class PhraseCorpus
 			
 		}
 		try{
-		r.close();
+			r.close();
 		}catch(IOException ioe){
 			ioe.printStackTrace();
 		}
diff --git a/gi/posterior-regularisation/prjava/src/phrase/Trainer.java b/gi/posterior-regularisation/prjava/src/phrase/Trainer.java
index ed7a6bbe..d1322c26 100644
--- a/gi/posterior-regularisation/prjava/src/phrase/Trainer.java
+++ b/gi/posterior-regularisation/prjava/src/phrase/Trainer.java
@@ -36,6 +36,8 @@ public class Trainer
         parser.accepts("skip-large-phrases").withRequiredArg().ofType(Integer.class).defaultsTo(5);
         parser.accepts("rare-word").withRequiredArg().ofType(Integer.class).defaultsTo(0);
         parser.accepts("rare-edge").withRequiredArg().ofType(Integer.class).defaultsTo(0);
+        parser.accepts("rare-phrase").withRequiredArg().ofType(Integer.class).defaultsTo(0);
+        parser.accepts("rare-context").withRequiredArg().ofType(Integer.class).defaultsTo(0);
         OptionSet options = parser.parse(args);
 
         if (options.has("help") || !options.has("in"))
@@ -61,6 +63,8 @@ public class Trainer
 		int skip = (Integer) options.valueOf("skip-large-phrases");
 		int wordThreshold = (Integer) options.valueOf("rare-word");
 		int edgeThreshold = (Integer) options.valueOf("rare-edge");
+		int phraseThreshold = (Integer) options.valueOf("rare-phrase");
+		int contextThreshold = (Integer) options.valueOf("rare-context");
 		
 		if (options.has("seed"))
 			F.rng = new Random((Long) options.valueOf("seed"));
@@ -83,8 +87,12 @@ public class Trainer
 			System.exit(1);
 		}
 		
-		if (wordThreshold > 0)
+		if (wordThreshold > 1)
 			corpus.applyWordThreshold(wordThreshold);
+		if (phraseThreshold > 1)
+			corpus.applyPhraseThreshold(phraseThreshold);
+		if (contextThreshold > 1)
+			corpus.applyContextThreshold(contextThreshold);
 		
 		if (!options.has("agree"))
 			System.out.println("Running with " + tags + " tags " +
-- 
cgit v1.2.3