From a942e3940c30eeba900c873f3bfb3f48d5b39ddb Mon Sep 17 00:00:00 2001
From: SimBe195 <37951951+SimBe195@users.noreply.github.com>
Date: Wed, 27 Mar 2024 11:13:14 +0100
Subject: [PATCH] Fix word boundaries in ctm dumps of lattices with subsampling
 (#72)

---
 src/Flf/Traceback.cc | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/src/Flf/Traceback.cc b/src/Flf/Traceback.cc
index 704c3697..289ae697 100644
--- a/src/Flf/Traceback.cc
+++ b/src/Flf/Traceback.cc
@@ -60,6 +60,7 @@ class DumpTracebackNode : public FilterNode {
     static const Core::ParameterBool         paramDumpPhonemeAlignment;
     static const Core::ParameterBool         paramDumpSubwordAlignment;
     static const Core::ParameterBool         paramFillEmptySegments;
+    static const Core::ParameterFloat        paramFrameShiftTime;
 
 private:
     Core::Channel                    dump_;
@@ -82,6 +83,7 @@ class DumpTracebackNode : public FilterNode {
     bool                       dumpPhonemeAlignment_;
     bool                       dumpSubwordAlignment_;
     bool                       fillEmptySegments_;
+    float                      frameShiftTime_;
     LatticeAlignmentBuilderRef alignmentBuilder_;
 
 protected:
@@ -370,8 +372,8 @@ class DumpTracebackNode : public FilterNode {
                     verify(sr->nArcs() == 1);
                     const Arc&      a            = *sr->begin();
                     const Boundary &leftBoundary = boundaries.get(sr->id()), &rightBoundary = boundaries.get(a.target());
-                    f32             wordBegin = f32(leftBoundary.time()) / 100.00;
-                    f32             wordEnd   = f32(rightBoundary.time()) / 100.00;
+                    f32             wordBegin = f32(leftBoundary.time()) * frameShiftTime_;
+                    f32             wordEnd   = f32(rightBoundary.time()) * frameShiftTime_;
                     if (wordBegin < wordEnd) {
                         if (lAlphabet || lpAlphabet) {
                             std::string      word;
@@ -580,6 +582,7 @@ class DumpTracebackNode : public FilterNode {
                 dumpPhonemeAlignment_               = paramDumpPhonemeAlignment(ctmConfig);
                 dumpSubwordAlignment_               = paramDumpSubwordAlignment(ctmConfig);
                 fillEmptySegments_                  = paramFillEmptySegments(ctmConfig);
+                frameShiftTime_                     = paramFrameShiftTime(ctmConfig);
                 if (dumpPhonemeAlignment_ || dumpSubwordAlignment_) {
                     createAlignmentBuilder(ctmConfig);
                 }
@@ -684,6 +687,10 @@ const Core::ParameterBool DumpTracebackNode::paramFillEmptySegments(
         "fill-empty-segments",
         "fill empty segments (can fix issues with sclite complaining about unsynchronized files if a segment is missing from the ctm file)",
         false);
+const Core::ParameterFloat DumpTracebackNode::paramFrameShiftTime(
+        "frame-shift-time",
+        "shift-time of frames of the lattice time axis in seconds. Defaults to 0.01 = 10ms. Important for correct word boundaries when subsampling is used.",
+        0.01);
 NodeRef createDumpTracebackNode(const std::string& name, const Core::Configuration& config) {
     return NodeRef(new DumpTracebackNode(name, config));
 }