From ec7edcc7e398bdb040d810094b8416ad9f279d98 Mon Sep 17 00:00:00 2001
From: Chris Dyer <redpony@gmail.com>
Date: Mon, 7 Dec 2009 01:50:49 -0500
Subject: minor changes to the way the phonotactic prob is calculated

---
 src/csplit.h | 7 +++++++
 1 file changed, 7 insertions(+)

(limited to 'src/csplit.h')

diff --git a/src/csplit.h b/src/csplit.h
index 5911af77..54e5329d 100644
--- a/src/csplit.h
+++ b/src/csplit.h
@@ -4,6 +4,13 @@
 #include "translator.h"
 #include "lattice.h"
 
+// this "translator" takes single words (with NO SPACES) and segments
+// them using the approach described in:
+//
+// C. Dyer. (2009) Using a maximum entropy model to build segmentation
+//                 lattices for MT. In Proceedings of NAACL HLT 2009.
+// note, an extra word space marker # is inserted at the left edge of
+// the forest!
 struct CompoundSplitImpl;
 struct CompoundSplit : public Translator {
   CompoundSplit(const boost::program_options::variables_map& conf);
-- 
cgit v1.2.3