svn commit: r1414249 - /stanbol/trunk/enhancer/engines/entitylinking/src/main/java/org/apache/stanbol/enhancer/engines/entitylinking/impl/EntityLinker.java

[rwesten]

Author: rwesten
Date: Tue Nov 27 16:24:38 2012
New Revision: 1414249

URL: http://svn.apache.org/viewvc?rev=1414249&view=rev
Log:
STANBOL-819, STANBOL-740: The EntityLinker now extends Queries also outside of 
Chunks. Tests have shown that this improves results. The old behaviour can be 
reactivated by a boolean switch. If further tests show that both options are 
useful depending on the domain, than one can make this also configureable

Modified:
    
stanbol/trunk/enhancer/engines/entitylinking/src/main/java/org/apache/stanbol/enhancer/engines/entitylinking/impl/EntityLinker.java

Modified: 
stanbol/trunk/enhancer/engines/entitylinking/src/main/java/org/apache/stanbol/enhancer/engines/entitylinking/impl/EntityLinker.java
URL: 
http://svn.apache.org/viewvc/stanbol/trunk/enhancer/engines/entitylinking/src/main/java/org/apache/stanbol/enhancer/engines/entitylinking/impl/EntityLinker.java?rev=1414249&r1=1414248&r2=1414249&view=diff
==============================================================================
--- 
stanbol/trunk/enhancer/engines/entitylinking/src/main/java/org/apache/stanbol/enhancer/engines/entitylinking/impl/EntityLinker.java
 (original)
+++ 
stanbol/trunk/enhancer/engines/entitylinking/src/main/java/org/apache/stanbol/enhancer/engines/entitylinking/impl/EntityLinker.java
 Tue Nov 27 16:24:38 2012
@@ -103,8 +103,13 @@ public class EntityLinker {
             searchStrings.add(token.getTokenText());
             //Determine the range we are allowed to search for tokens
             final int minIncludeIndex;
-            int maxIndcludeIndex;
-            if(token.inChunk != null && 
!textProcessingConfig.isIgnoreChunks()){
+            final int maxIndcludeIndex;
+            //NOTE: testing has shown that using Chunks to restrict search for
+            //      additional matchable tokens does have an negative impact on
+            //      recall. Because of that this restriction is for now 
deactivated
+            boolean restrirctContextByChunks = false; //TODO: maybe make 
configurable
+            if(token.inChunk != null && !textProcessingConfig.isIgnoreChunks() 
&&
+                    restrirctContextByChunks){
                 minIncludeIndex = Math.max(
                     state.getConsumedIndex()+1, 
                     token.inChunk.startToken);