Update lexicon phone to pronunciations (#13)

CMakeLists.txt

@@ -39,7 +39,7 @@ if (MSVC)
      set(CMAKE_INSTALL_RPATH "$ORIGIN;$ORIGIN/../lib;$ORIGIN/../../tools/openfst/lib")
 endif ()
 
-find_package(CUDA)
+find_package(CUDAToolkit)
 
 find_package(pybind11 REQUIRED)
 include_directories(extensions)
@@ -86,7 +86,7 @@ target_link_libraries(_kalpy PUBLIC kaldi-base kaldi-chain
                                     fstscript
                                     )
 
-if(CUDA_FOUND)
+if(CUDAToolkit_FOUND)
 
    target_link_libraries(_kalpy PUBLIC kaldi-cudadecoder kaldi-cudafeat
                                     )

docs/source/changelog.rst

@@ -4,6 +4,11 @@
 Changelog
 =========
 
+0.6.0
+-----
+
+- Fixed a bug in feature archives where fMLLR transforms were being ignored
+
 0.5.1
 -----
 

kalpy/fstext/lexicon.py

@@ -916,10 +916,10 @@ def _create_pronunciation_string(
 
     def phones_to_pronunciations(
         self,
-        text: str,
         word_symbols: typing.List[int],
         intervals: typing.List[CtmInterval],
         transcription: bool = False,
+        text: str = None,
     ) -> HierarchicalCtm:
 
         phones = [x.symbol for x in intervals]
@@ -928,12 +928,10 @@ def phones_to_pronunciations(
             phones,
             transcription=transcription,
         )
-        if transcription:
-            actual_words = [self.word_table.find(x) for x in word_symbols]
-            if not text:
-                text = " ".join(actual_words)
-        else:
-            actual_words = [x for x in text.split() if x != self.silence_word]
+
+        actual_words = [self.word_table.find(x) for x in word_symbols]
+        if not text:
+            text = " ".join(actual_words)
         word_intervals = []
         current_phone_index = 0
         current_word_index = 0
@@ -1002,10 +1000,10 @@ def __init__(
 
     def phones_to_pronunciations(
         self,
-        text: str,
         word_symbols: typing.List[int],
         intervals: typing.List[CtmInterval],
         transcription: bool = False,
+        text: str = None,
     ) -> HierarchicalCtm:
         phone_symbols = [x.symbol for x in intervals]
         word_symbols = [self.word_table.find(x) for x in text.split()]

tests/test_align.py

@@ -76,8 +76,7 @@ def test_align_sat_first_pass(
     alignment = alignment_archive["1-1"]
     assert len(alignment.alignment) == 2672
     intervals = alignment.generate_ctm(aligner.transition_model, lc.phone_table)
-    text = " ".join(lc.word_table.find(x) for x in alignment.words)
-    ctm = lc.phones_to_pronunciations(text, alignment.words, intervals)
+    ctm = lc.phones_to_pronunciations(alignment.words, intervals)
     ctm.export_textgrid(textgrid_name, file_duration=26.72)
     reference_alignment_archive = AlignmentArchive(reference_first_pass_ali_path)
     reference_alignment = reference_alignment_archive["1-1"]
@@ -129,8 +128,7 @@ def test_align_sat_second_pass(
     alignment = alignment_archive["1-1"]
     assert len(alignment.alignment) == 2672
     intervals = alignment.generate_ctm(aligner.transition_model, lc.phone_table)
-    text = " ".join(lc.word_table.find(x) for x in alignment.words)
-    ctm = lc.phones_to_pronunciations(text, alignment.words, intervals)
+    ctm = lc.phones_to_pronunciations(alignment.words, intervals)
     ctm.export_textgrid(textgrid_name, file_duration=26.72)
 
     reference_alignment_archive = AlignmentArchive(reference_second_pass_ali_path)

tests/test_decode.py

@@ -77,8 +77,7 @@ def test_decode_sat_first_pass(
     alignment = alignment_archive["1-1"]
     assert len(alignment.alignment) == 2672
     intervals = alignment.generate_ctm(decoder.transition_model, lc.phone_table)
-    text = " ".join(lc.word_table.find(x) for x in alignment.words)
-    ctm = lc.phones_to_pronunciations(text, alignment.words, intervals)
+    ctm = lc.phones_to_pronunciations(alignment.words, intervals)
     ctm.export_textgrid(textgrid_name, file_duration=26.72)
 
 
@@ -128,8 +127,7 @@ def test_decode_sat_second_pass(
     alignment = alignment_archive["1-1"]
     assert len(alignment.alignment) == 2672
     intervals = alignment.generate_ctm(decoder.transition_model, lc.phone_table)
-    text = " ".join(lc.word_table.find(x) for x in alignment.words)
-    ctm = lc.phones_to_pronunciations(text, alignment.words, intervals)
+    ctm = lc.phones_to_pronunciations(alignment.words, intervals)
     ctm.export_textgrid(textgrid_name, file_duration=26.72)