Added Cardinal class for Telugu TN

Jenkinsfile

@@ -28,6 +28,9 @@ pipeline {
     HY_TN_CACHE='/home/jenkins/TestData/text_norm/ci/grammars/03-12-24-0'
     MR_TN_CACHE='/home/jenkins/TestData/text_norm/ci/grammars/03-12-24-1'
     JA_TN_CACHE='/home/jenkins/TestData/text_norm/ci/grammars/10-17-24-1'
+    KO_TN_CACHE='/home/jenkins/TestData/text_norm/ci/grammars/04-23-26-0'
+    HI_TN_CACHE='/home/jenkins/TestData/text_norm/ci/grammars/04-23-26-0'
+    TE_TN_CACHE='/home/jenkins/TestData/text_norm/ci/grammars/07-21-26-0'
     HI_TN_CACHE='/home/jenkins/TestData/text_norm/ci/grammars/06-04-26-5'
     KO_TN_CACHE='/home/jenkins/TestData/text_norm/ci/grammars/06-04-25-6'
     DEFAULT_TN_CACHE='/home/jenkins/TestData/text_norm/ci/grammars/06-08-23-0'
@@ -112,6 +115,30 @@ pipeline {
         }
       }
     }
+        stage('L0: Create TE TN/ITN Grammars') {
+    when {
+        anyOf {
+            branch 'main'
+            branch 'staging/**'
+            branch 'staging_*'
+            changeRequest target: 'main'
+        }
+    }
+    failFast true
+    parallel {
+        stage('L0: TE TN grammars') {
+            steps {
+                sh 'CUDA_VISIBLE_DEVICES="" python nemo_text_processing/text_normalization/normalize.py --lang=te --text="౧" --cache_dir ${TE_TN_CACHE}'
+            }
+        }
+        // stage('L0: TE ITN grammars') {
+        //     steps {
+        //         sh 'CUDA_VISIBLE_DEVICES="" python nemo_text_processing/inverse_text_normalization/inverse_normalize.py --lang=te --text="ఒకటి" --cache_dir ${TE_TN_CACHE}'
+        //     }
+        // }
+
+      }
+    }
 
     stage('L0: Create DE/ES TN/ITN Grammars') {
       when {
@@ -407,6 +434,11 @@ pipeline {
           steps {
             sh 'CUDA_VISIBLE_DEVICES="" pytest tests/nemo_text_processing/hi/ -m "not pleasefixme" --cpu --tn_cache_dir ${HI_TN_CACHE}'
           }
+        }
+         stage('L1: Run all TE TN/ITN tests (restore grammars from cache)') {
+          steps {
+            sh 'CUDA_VISIBLE_DEVICES="" pytest tests/nemo_text_processing/te/ -m "not pleasefixme" --cpu --tn_cache_dir ${TE_TN_CACHE}'
+          }
         }
         stage('L1: Run all Codeswitched ES/EN TN/ITN tests (restore grammars from cache)') {
           steps {

nemo_text_processing/text_normalization/normalize.py

@@ -166,6 +166,7 @@ def __init__(
 
             if post_process:
                 self.post_processor = PostProcessingFst(cache_dir=cache_dir, overwrite_cache=overwrite_cache)
+
         elif lang == 'it':
             from nemo_text_processing.text_normalization.it.taggers.tokenize_and_classify import ClassifyFst
             from nemo_text_processing.text_normalization.it.verbalizers.verbalize_final import VerbalizeFinalFst
@@ -191,6 +192,13 @@ def __init__(
         elif lang == 'ko':
             from nemo_text_processing.text_normalization.ko.taggers.tokenize_and_classify import ClassifyFst
             from nemo_text_processing.text_normalization.ko.verbalizers.verbalize_final import VerbalizeFinalFst
+        elif lang == 'te':
+            from nemo_text_processing.text_normalization.te.taggers.tokenize_and_classify import ClassifyFst
+            from nemo_text_processing.text_normalization.te.verbalizers.post_processing import PostProcessingFst
+            from nemo_text_processing.text_normalization.te.verbalizers.verbalize_final import VerbalizeFinalFst
+
+            if post_process:
+                self.post_processor = PostProcessingFst(cache_dir=cache_dir, overwrite_cache=overwrite_cache)
         else:
             raise NotImplementedError(f"Language {lang} has not been supported yet.")
 
@@ -391,7 +399,11 @@ def normalize(
                 return text
         output = SPACE_DUP.sub(' ', output[1:])
 
-        if self.lang in ["en", "hi", "vi"] and hasattr(self, 'post_processor') and self.post_processor is not None:
+        if (
+            self.lang in ["en", "hi", "te", "vi"]
+            and hasattr(self, 'post_processor')
+            and self.post_processor is not None
+        ):
             output = self.post_process(output)
 
         if punct_post_process:
@@ -737,7 +749,7 @@ def parse_args():
     parser.add_argument(
         "--language",
         help="language",
-        choices=["en", "de", "es", "fr", "hu", "sv", "zh", "ar", "it", "hy", "ja", "hi", "ko", "vi", "pt"],
+        choices=["en", "de", "es", "fr", "hu", "sv", "zh", "ar", "it", "hy", "ja", "hi", "te", "ko", "vi", "pt"],
         default="en",
         type=str,
     )

nemo_text_processing/text_normalization/run_evaluate.py

@@ -35,7 +35,7 @@ def parse_args():
     parser.add_argument(
         "--lang",
         help="language",
-        choices=['ar', 'de', 'en', 'es', 'fr', 'hu', 'it', 'ru', 'sv', 'zh', 'hy', 'hi', 'ko', 'vi', 'pt'],
+        choices=['ar', 'de', 'en', 'es', 'fr', 'hu', 'it', 'ru', 'sv', 'zh', 'hy', 'hi', 'te', 'ko', 'vi', 'pt'],
         default="en",
         type=str,
     )

nemo_text_processing/text_normalization/te/__init__.py

@@ -0,0 +1,13 @@
+# Copyright (c) 2026, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.

nemo_text_processing/text_normalization/te/data/numbers/__init__.py

@@ -0,0 +1,13 @@
+# Copyright (c) 2026, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.

nemo_text_processing/text_normalization/te/data/numbers/digit.tsv

@@ -0,0 +1,18 @@
+1	ఒకటి
+2	రెండు
+3	మూడు
+4	నాలుగు
+5	ఐదు
+6	ఆరు
+7	ఏడు
+8	ఎనిమిది
+9	తొమ్మిది
+౧	ఒకటి
+౨	రెండు
+౩	మూడు
+౪	నాలుగు
+౫	ఐదు
+౬	ఆరు
+౭	ఏడు
+౮	ఎనిమిది
+౯	తొమ్మిది

nemo_text_processing/text_normalization/te/data/numbers/magnitudes.tsv

@@ -0,0 +1,15 @@
+hundred	వంద
+hundred_prefix	నూట
+hundreds_plural	వందలు
+hundreds_before	వందల
+thousand	వెయ్యి
+thousands_plural	వేలు
+thousands_before	వేల
+lakh	లక్ష
+lakh_before_digit	లక్షా
+lakhs_plural	లక్షలు
+lakhs_before	లక్షల
+crore	కోటి
+crores_plural	కోట్లు
+crores_before	కోట్ల
+one	ఒక

nemo_text_processing/text_normalization/te/data/numbers/teens_and_ties.tsv

@@ -0,0 +1,10 @@
+౧౦	పది
+౧౧	పదకొండు
+౧౨	పన్నెండు
+౧౩	పదమూడు
+౧౪	పద్నాలుగు
+౧౫	పదిహేను
+౧౬	పదహారు
+౧౭	పదిహేడు
+౧౮	పధ్ధెనిమిది
+౧౯	పంతొమ్మిది

nemo_text_processing/text_normalization/te/data/numbers/teens_and_ties_en.tsv

@@ -0,0 +1,10 @@
+10	పది
+11	పదకొండు
+12	పన్నెండు
+13	పదమూడు
+14	పద్నాలుగు
+15	పదిహేను
+16	పదహారు
+17	పదిహేడు
+18	పధ్ధెనిమిది
+19	పంతొమ్మిది

nemo_text_processing/text_normalization/te/data/numbers/teens_and_ties_thousand.tsv

@@ -0,0 +1,16 @@
+21	ఇరవై ఒక
+31	ముప్పై ఒక
+41	నలభై ఒక
+51	యాభై ఒక
+61	అరవై ఒక
+71	డెబ్బై ఒక
+81	ఎనభై ఒక
+91	తొంభై ఒక
+౨౧	ఇరవై ఒక
+౩౧	ముప్పై ఒక
+౪౧	నలభై ఒక
+౫౧	యాభై ఒక
+౬౧	అరవై ఒక
+౭౧	డెబ్బై ఒక
+౮౧	ఎనభై ఒక
+౯౧	తొంభై ఒక

nemo_text_processing/text_normalization/te/data/numbers/ties.tsv

@@ -0,0 +1,16 @@
+2	ఇరవై
+3	ముప్పై
+4	నలభై
+5	యాభై
+6	అరవై
+7	డెబ్బై
+8	ఎనభై
+9	తొంభై
+౨	ఇరవై
+౩	ముప్పై
+౪	నలభై
+౫	యాభై
+౬	అరవై
+౭	డెబ్బై
+౮	ఎనభై
+౯	తొంభై

nemo_text_processing/text_normalization/te/data/numbers/zero.tsv

@@ -0,0 +1,2 @@
+0	సున్నా
+౦	సున్నా

nemo_text_processing/text_normalization/te/graph_utils.py

@@ -0,0 +1,145 @@
+# Copyright (c) 2026, NVIDIA CORPORATION.  All rights reserved.
+# Copyright 2015 and onwards Google, Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import logging
+import os
+from pathlib import Path
+from typing import Dict
+
+import pynini
+from pynini import Far
+from pynini.export import export
+from pynini.lib import byte, pynutil, utf8
+
+NEMO_CHAR = utf8.VALID_UTF8_CHAR
+NEMO_DIGIT = byte.DIGIT
+
+# Telugu native digits (౦, ౧, ౨, ౩, ౪, ౫, ౬, ౭, ౮, ౯)
+NEMO_TE_DIGIT = pynini.union("౦", "౧", "౨", "౩", "౪", "౫", "౬", "౭", "౮", "౯").optimize()
+
+# Combined Telugu and Arabic digits for graphs that need to accept both
+NEMO_ALL_DIGIT = pynini.union(NEMO_TE_DIGIT, NEMO_DIGIT).optimize()
+NEMO_ALL_ZERO = pynini.union("౦", "0").optimize()
+
+NEMO_NON_BREAKING_SPACE = u"\u00a0"
+NEMO_SPACE = " "
+NEMO_WHITE_SPACE = pynini.union(" ", "\t", "\n", "\r", u"\u00a0").optimize()
+NEMO_NOT_SPACE = pynini.difference(NEMO_CHAR, NEMO_WHITE_SPACE).optimize()
+NEMO_NOT_QUOTE = pynini.difference(NEMO_CHAR, r'"').optimize()
+NEMO_SIGMA = pynini.closure(NEMO_CHAR)
+
+delete_space = pynutil.delete(pynini.closure(NEMO_WHITE_SPACE))
+insert_space = pynutil.insert(" ")
+delete_extra_space = pynini.cross(pynini.closure(NEMO_WHITE_SPACE, 1), " ")
+
+MIN_NEG_WEIGHT = -0.0001
+
+
+def generator_main(file_name: str, graphs: Dict[str, 'pynini.FstLike']):
+    """
+    Exports graph as OpenFst finite state archive (FAR) file with given file name and rule name.
+
+    Args:
+        file_name: exported file name
+        graphs: Mapping of a rule name and Pynini WFST graph to be exported
+    """
+    exporter = export.Exporter(file_name)
+    for rule, graph in graphs.items():
+        exporter[rule] = graph.optimize()
+    exporter.close()
+    logging.info(f'Created {file_name}')
+
+
+def convert_space(fst) -> 'pynini.FstLike':
+    """
+    Converts space to nonbreaking space.
+    Used only in tagger grammars for transducing token values within quotes, e.g. name: "hello kitty"
+    This is making transducer significantly slower, so only use when there could be potential spaces within quotes, otherwise leave it.
+
+    Args:
+        fst: input fst
+
+    Returns output fst where breaking spaces are converted to non breaking spaces
+    """
+    return fst @ pynini.cdrewrite(pynini.cross(NEMO_SPACE, NEMO_NON_BREAKING_SPACE), "", "", NEMO_SIGMA)
+
+
+class GraphFst:
+    """
+    Base class for all grammar fsts.
+
+    Args:
+        name: name of grammar class
+        kind: either 'classify' or 'verbalize'
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, name: str, kind: str, deterministic: bool = True):
+        self.name = name
+        self.kind = kind
+        self._fst = None
+        self.deterministic = deterministic
+
+        self.far_path = Path(os.path.dirname(__file__) + '/grammars/' + kind + '/' + name + '.far')
+        if self.far_exist():
+            self._fst = Far(self.far_path, mode="r", arc_type="standard", far_type="default").get_fst()
+
+    def far_exist(self) -> bool:
+        """
+        Returns true if FAR can be loaded
+        """
+        return self.far_path.exists()
+
+    @property
+    def fst(self) -> 'pynini.FstLike':
+        return self._fst
+
+    @fst.setter
+    def fst(self, fst):
+        self._fst = fst
+
+    def add_tokens(self, fst) -> 'pynini.FstLike':
+        """
+        Wraps class name around to given fst
+
+        Args:
+            fst: input fst
+
+        Returns:
+            Fst: fst
+        """
+        return pynutil.insert(f"{self.name} {{ ") + fst + pynutil.insert(" }")
+
+    def delete_tokens(self, fst) -> 'pynini.FstLike':
+        """
+        Deletes class name wrap around output of given fst
+
+        Args:
+            fst: input fst
+
+        Returns:
+            Fst: fst
+        """
+        res = (
+            pynutil.delete(f"{self.name}")
+            + delete_space
+            + pynutil.delete("{")
+            + delete_space
+            + fst
+            + delete_space
+            + pynutil.delete("}")
+        )
+        return res @ pynini.cdrewrite(pynini.cross(u"\u00a0", " "), "", "", NEMO_SIGMA)

nemo_text_processing/text_normalization/te/taggers/__init__.py

@@ -0,0 +1,13 @@
+# Copyright (c) 2026, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.