Revision: 16656
http://sourceforge.net/p/gate/code/16656
Author: adamfunk
Date: 2013-04-23 12:44:16 +0000 (Tue, 23 Apr 2013)
Log Message:
-----------
Conll format: adding support for L-TAG & U-TAG.
Modified Paths:
--------------
gate/trunk/src/gate/corpora/ConllDocumentFormat.java
Modified: gate/trunk/src/gate/corpora/ConllDocumentFormat.java
===================================================================
--- gate/trunk/src/gate/corpora/ConllDocumentFormat.java 2013-04-23
01:50:23 UTC (rev 16655)
+++ gate/trunk/src/gate/corpora/ConllDocumentFormat.java 2013-04-23
12:44:16 UTC (rev 16656)
@@ -112,6 +112,33 @@
finishAllTags(inProgress, annotanda, oldEnd);
}
+ // "U-FOO": unigram, single-token "FOO"
+ // annotation, after closing any "FOO" already in progress
+ else if ( (items[column].length() > 2) &&
+ items[column].startsWith("U-") ) {
+ String type = items[column].substring(2);
+ finishTag(type, inProgress, annotanda, oldEnd);
+ annotanda.add(new Annotandum(type, start, end, column, true));
+ }
+
+ // "L-FOO": last bit of "FOO": extend and
+ // close any "FOO" already in progress
+ else if ( (items[column].length() > 2) &&
+ items[column].startsWith("L-") ) {
+ String type = items[column].substring(2);
+
+ if (inProgress.containsKey(type)) {
+ // good L-FOO, so update the end offset
+ inProgress.get(type).endOffset = end;
+ }
+ else {
+ // bad data, containing I-FOO without a B-FOO, so treat as if
B-FOO
+ inProgress.put(type, new Annotandum(type, start, end, column,
true));
+ }
+
+ finishTag(type, inProgress, annotanda, end);
+ }
+
// "B-FOO": start a new "FOO" annotation
// after closing any "FOO" already in progress
else if ( (items[column].length() > 2) &&
@@ -138,7 +165,7 @@
// "FOO": treat as single-token annotation (such as POS tag)
else {
- Annotandum tag = new Annotandum(items[1], start, end, column,
false);
+ Annotandum tag = new Annotandum(items[column], start, end, column,
false);
annotanda.add(tag);
}
}
@@ -236,7 +263,7 @@
}
/* Note that chunkiness is determined by the tag structure. A "B-Foo"
- * that spans only one token is chunky. Tags outside the B/I/O system
+ * that spans only one token is chunky. Tags outside the B/I/L/U system
* get the kind==token feature; tags in the system get kind==chunky. */
protected Annotandum(String type, Long startOffset, Long endOffset, int
column, boolean chunky) {
this.features = Factory.newFeatureMap();
This was sent by the SourceForge.net collaborative development platform, the
world's largest Open Source development site.
------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_apr
_______________________________________________
GATE-cvs mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/gate-cvs
