chore: import upstream snapshot with attribution

2026-07-13 13:25:10 +08:00
commit c397331b1e
3684 changed files with 990993 additions and 0 deletions
@@ -0,0 +1 @@
+
@@ -0,0 +1 @@
+
@@ -0,0 +1,31 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import DAMO_DIGIT, GraphFst, insert_space
+from pynini.lib import pynutil
+
+
+class CardinalFst(GraphFst):
+    """
+    Finite state transducer for classifying cardinals, e.g.
+       "тысяча один" ->  cardinal { integer: "1 001" }
+
+    Args:
+        tn_cardinal: Text normalization Cardinal graph
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, tn_cardinal: GraphFst, deterministic: bool = True):
+        super().__init__(name="cardinal", kind="classify", deterministic=deterministic)
+
+        graph = tn_cardinal.cardinal_numbers_default
+        self.graph = graph.invert().optimize()
+
+        optional_sign = pynini.closure(
+            pynutil.insert("negative: ") + pynini.cross("минус ", '"-"') + insert_space, 0, 1
+        )
+
+        # do not invert numbers less than 10
+        graph = pynini.compose(graph, DAMO_DIGIT ** (2, ...))
+        graph = optional_sign + pynutil.insert('integer: "') + graph + pynutil.insert('"')
+        graph = self.add_tokens(graph)
+        self.fst = graph.optimize()
@@ -0,0 +1,22 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import GraphFst
+from pynini.lib import pynutil
+
+
+class DateFst(GraphFst):
+    """
+    Finite state transducer for classifying date, e.g.
+        восемнадцатое июня две тысячи второго -> tokens { date { day: "18.06.2002" } }
+
+    Args:
+        tn_date: Text normalization Date graph
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, tn_date: GraphFst, deterministic: bool = True):
+        super().__init__(name="date", kind="classify", deterministic=deterministic)
+
+        graph = pynini.invert(tn_date.final_graph).optimize()
+        graph = self.add_tokens(pynutil.insert('day: "') + graph + pynutil.insert('"'))
+        self.fst = graph.optimize()
@@ -0,0 +1,51 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import (
+    DAMO_SPACE,
+    GraphFst,
+    delete_extra_space,
+)
+from pynini.lib import pynutil
+
+
+class DecimalFst(GraphFst):
+    """
+    Finite state transducer for classifying decimal
+        e.g. "минус три целых две десятых" -> decimal { negative: "true" integer_part: "3," fractional_part: "2" }
+
+    Args:
+        tn_decimal: Text normalization Decimal graph
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, tn_decimal, deterministic: bool = False):
+        super().__init__(name="decimal", kind="classify", deterministic=deterministic)
+
+        optional_graph_negative = pynini.closure(
+            pynutil.insert("negative: ") + pynini.cross("минус", '"true"') + delete_extra_space,
+            0,
+            1,
+        )
+
+        graph_fractional_part = pynini.invert(tn_decimal.graph_fractional).optimize()
+        graph_integer_part = pynini.invert(tn_decimal.integer_part).optimize()
+        optional_graph_quantity = pynini.invert(tn_decimal.optional_quantity).optimize()
+
+        graph_fractional = (
+            pynutil.insert('fractional_part: "') + graph_fractional_part + pynutil.insert('"')
+        )
+        graph_integer = pynutil.insert('integer_part: "') + graph_integer_part + pynutil.insert('"')
+        optional_graph_quantity = (
+            pynutil.insert('quantity: "') + optional_graph_quantity + pynutil.insert('"')
+        )
+        optional_graph_quantity = pynini.closure(
+            pynini.accep(DAMO_SPACE) + optional_graph_quantity, 0, 1
+        )
+
+        self.final_graph_wo_sign = (
+            graph_integer + pynini.accep(DAMO_SPACE) + graph_fractional + optional_graph_quantity
+        )
+        final_graph = optional_graph_negative + self.final_graph_wo_sign
+
+        final_graph = self.add_tokens(final_graph)
+        self.fst = final_graph.optimize()
@@ -0,0 +1,23 @@
+from fun_text_processing.text_normalization.en.graph_utils import GraphFst
+from pynini.lib import pynutil
+
+
+class ElectronicFst(GraphFst):
+    """
+    Finite state transducer for classifying electronic, e.g.
+        "эй би собака эн ди точка ру" -> electronic { username: "ab@nd.ru" }
+
+    Args:
+        tn_electronic: Text normalization Electronic graph
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, tn_electronic, deterministic: bool = True):
+        super().__init__(name="electronic", kind="classify", deterministic=deterministic)
+
+        graph = tn_electronic.final_graph
+        graph = graph.invert().optimize()
+        graph = pynutil.insert('username: "') + graph + pynutil.insert('"')
+        graph = self.add_tokens(graph)
+        self.fst = graph.optimize()
@@ -0,0 +1,23 @@
+from fun_text_processing.text_normalization.en.graph_utils import GraphFst
+from pynini.lib import pynutil
+
+
+class MeasureFst(GraphFst):
+    """
+    Finite state transducer for classifying measure, e.g.
+        "два килограма" -> measure { cardinal { integer: "2 кг" } }
+
+    Args:
+        tn_measure: Text normalization Cardinal graph
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, tn_measure, deterministic: bool = True):
+        super().__init__(name="measure", kind="classify", deterministic=deterministic)
+
+        tn_measure = tn_measure.tagger_graph_default @ tn_measure.verbalizer_graph
+        graph = tn_measure.invert().optimize()
+        graph = pynutil.insert('cardinal { integer: "') + graph + pynutil.insert('" }')
+        graph = self.add_tokens(graph)
+        self.fst = graph.optimize()
@@ -0,0 +1,23 @@
+from fun_text_processing.text_normalization.en.graph_utils import GraphFst
+from pynini.lib import pynutil
+
+
+class MoneyFst(GraphFst):
+    """
+    Finite state transducer for classifying money, e.g.
+        "два рубля" -> money { integer_part: "2 руб." }
+
+    Args:
+        tn_money: Text normalization Money graph
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, tn_money, deterministic: bool = True):
+        super().__init__(name="money", kind="classify", deterministic=deterministic)
+
+        graph = tn_money.final_graph
+        graph = graph.invert().optimize()
+        graph = pynutil.insert('integer_part: "') + graph + pynutil.insert('"')
+        graph = self.add_tokens(graph)
+        self.fst = graph.optimize()
@@ -0,0 +1,29 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import DAMO_DIGIT, GraphFst
+from pynini.lib import pynutil
+
+
+class OrdinalFst(GraphFst):
+    """
+    Finite state transducer for classifying ordinals, e.g.
+        "второе" -> ordinal { integer: "2" } }
+
+    Args:
+        tn_ordinal: Text normalization Ordinal graph
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, tn_ordinal: GraphFst, deterministic: bool = True):
+        super().__init__(name="ordinal", kind="classify", deterministic=deterministic)
+
+        tn_ordinal = tn_ordinal.ordinal_numbers
+
+        graph = tn_ordinal.invert().optimize()
+        self.graph = graph
+
+        # do not invert numbers less than 10
+        graph = pynini.compose(graph, DAMO_DIGIT ** (2, ...))
+        graph = pynutil.insert('integer: "') + graph + pynutil.insert('"')
+        graph = self.add_tokens(graph)
+        self.fst = graph.optimize()
@@ -0,0 +1,23 @@
+from fun_text_processing.text_normalization.en.graph_utils import GraphFst
+from pynini.lib import pynutil
+
+
+class TelephoneFst(GraphFst):
+    """
+    Finite state transducer for classifying telephone, e.g.
+        "восемь девятьсот тринадцать девятьсот восемьдесят три пятьдесят шесть ноль один" -> telephone { number_part: "8-913-983-56-01" }
+
+    Args:
+        tn_telephone: Text normalization telephone graph
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, tn_telephone: GraphFst, deterministic: bool = True):
+        super().__init__(name="telephone", kind="classify", deterministic=deterministic)
+
+        tn_telephone = tn_telephone.final_graph
+        graph = tn_telephone.invert().optimize()
+        graph = pynutil.insert('number_part: "') + graph + pynutil.insert('"')
+        graph = self.add_tokens(graph)
+        self.fst = graph.optimize()
@@ -0,0 +1,56 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import DAMO_SPACE, GraphFst
+from fun_text_processing.text_normalization.ru.verbalizers.time import TimeFst as TNTimeVerbalizer
+from pynini.lib import pynutil
+
+
+class TimeFst(GraphFst):
+    """
+    Finite state transducer for classifying time, e.g.
+        "два часа пятнадцать минут" -> time { hours: "02:15" }
+
+    Args:
+        tn_time: Text Normalization Time graph
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, tn_time: GraphFst, deterministic: bool = True):
+        super().__init__(name="time", kind="classify", deterministic=deterministic)
+
+        tn_time_tagger = tn_time.graph_preserve_order
+        tn_time_verbalizer = TNTimeVerbalizer().graph
+        tn_time_graph_preserve_order = pynini.compose(tn_time_tagger, tn_time_verbalizer).optimize()
+        graph_preserve_order = pynini.invert(tn_time_graph_preserve_order).optimize()
+        graph_preserve_order = (
+            pynutil.insert('hours: "') + graph_preserve_order + pynutil.insert('"')
+        )
+
+        # "пятнадцать минут шестого" -> 17:15
+        # Requires permutations for the correct verbalization
+        m_next_h = (
+            pynutil.insert('minutes: "')
+            + pynini.invert(tn_time.minutes).optimize()
+            + pynutil.insert('"')
+            + pynini.accep(DAMO_SPACE)
+            + pynutil.insert('hours: "')
+            + pynini.invert(tn_time.increment_hour_ordinal).optimize()
+            + pynutil.insert('"')
+        ).optimize()
+
+        # "без пятнадцати минут шесть" -> 17:45
+        # Requires permutation for the correct verbalization
+        m_to_h = (
+            pynini.cross("без ", 'minutes: "')
+            + pynini.invert(tn_time.mins_to_h)
+            + pynutil.insert('"')
+            + pynini.accep(DAMO_SPACE)
+            + pynutil.insert('hours: "')
+            + pynini.invert(tn_time.increment_hour_cardinal).optimize()
+            + pynutil.insert('"')
+        )
+
+        graph_reserve_order = m_next_h | m_to_h
+        graph = graph_preserve_order | graph_reserve_order
+        graph = self.add_tokens(graph)
+        self.fst = graph.optimize()
@@ -0,0 +1,110 @@
+import os
+
+import pynini
+from fun_text_processing.inverse_text_normalization.en.taggers.punctuation import PunctuationFst
+from fun_text_processing.inverse_text_normalization.en.taggers.word import WordFst
+from fun_text_processing.inverse_text_normalization.ru.taggers.cardinal import CardinalFst
+from fun_text_processing.inverse_text_normalization.ru.taggers.date import DateFst
+from fun_text_processing.inverse_text_normalization.ru.taggers.decimals import DecimalFst
+from fun_text_processing.inverse_text_normalization.ru.taggers.electronic import ElectronicFst
+from fun_text_processing.inverse_text_normalization.ru.taggers.measure import MeasureFst
+from fun_text_processing.inverse_text_normalization.ru.taggers.money import MoneyFst
+from fun_text_processing.inverse_text_normalization.ru.taggers.ordinal import OrdinalFst
+from fun_text_processing.inverse_text_normalization.ru.taggers.telephone import TelephoneFst
+from fun_text_processing.inverse_text_normalization.ru.taggers.time import TimeFst
+from fun_text_processing.inverse_text_normalization.ru.taggers.whitelist import WhiteListFst
+from fun_text_processing.text_normalization.en.graph_utils import (
+    GraphFst,
+    delete_extra_space,
+    delete_space,
+    generator_main,
+)
+from fun_text_processing.text_normalization.ru.taggers.tokenize_and_classify import (
+    ClassifyFst as TNClassifyFst,
+)
+from pynini.lib import pynutil
+
+import logging
+
+
+class ClassifyFst(GraphFst):
+    """
+    Final class that composes all other classification grammars. This class can process an entire sentence, that is lower cased.
+    For deployment, this grammar will be compiled and exported to OpenFst Finate State Archiv (FAR) File.
+    More details to deployment at NeMo/tools/text_processing_deployment.
+
+    Args:
+        cache_dir: path to a dir with .far grammar file. Set to None to avoid using cache.
+        overwrite_cache: set to True to overwrite .far files
+    """
+
+    def __init__(self, cache_dir: str = None, overwrite_cache: bool = False):
+        super().__init__(name="tokenize_and_classify", kind="classify")
+
+        far_file = None
+        if cache_dir is not None and cache_dir != "None":
+            os.makedirs(cache_dir, exist_ok=True)
+            far_file = os.path.join(cache_dir, "_ru_itn.far")
+        if not overwrite_cache and far_file and os.path.exists(far_file):
+            self.fst = pynini.Far(far_file, mode="r")["tokenize_and_classify"]
+            logging.info(f"ClassifyFst.fst was restored from {far_file}.")
+        else:
+            logging.info(f"Creating ClassifyFst grammars. This might take some time...")
+            tn_classify = TNClassifyFst(
+                input_case="cased", deterministic=False, cache_dir=cache_dir, overwrite_cache=True
+            )
+            cardinal = CardinalFst(tn_cardinal=tn_classify.cardinal)
+            cardinal_graph = cardinal.fst
+
+            ordinal = OrdinalFst(tn_ordinal=tn_classify.ordinal)
+            ordinal_graph = ordinal.fst
+
+            decimal = DecimalFst(tn_decimal=tn_classify.decimal)
+            decimal_graph = decimal.fst
+
+            measure_graph = MeasureFst(tn_measure=tn_classify.measure).fst
+            date_graph = DateFst(tn_date=tn_classify.date).fst
+            word_graph = WordFst().fst
+            time_graph = TimeFst(tn_time=tn_classify.time).fst
+            money_graph = MoneyFst(tn_money=tn_classify.money).fst
+            whitelist_graph = WhiteListFst().fst
+            punct_graph = PunctuationFst().fst
+            electronic_graph = ElectronicFst(tn_electronic=tn_classify.electronic).fst
+            telephone_graph = TelephoneFst(tn_telephone=tn_classify.telephone).fst
+
+            classify = (
+                pynutil.add_weight(whitelist_graph, 1.01)
+                | pynutil.add_weight(time_graph, 1.1)
+                | pynutil.add_weight(date_graph, 1.09)
+                | pynutil.add_weight(decimal_graph, 1.1)
+                | pynutil.add_weight(measure_graph, 1.1)
+                | pynutil.add_weight(ordinal_graph, 1.1)
+                | pynutil.add_weight(money_graph, 1.1)
+                | pynutil.add_weight(telephone_graph, 1.1)
+                | pynutil.add_weight(electronic_graph, 1.1)
+                | pynutil.add_weight(cardinal_graph, 1.1)
+                | pynutil.add_weight(word_graph, 100)
+            )
+
+            punct = (
+                pynutil.insert("tokens { ")
+                + pynutil.add_weight(punct_graph, weight=1.1)
+                + pynutil.insert(" }")
+            )
+            token = pynutil.insert("tokens { ") + classify + pynutil.insert(" }")
+            token_plus_punct = (
+                pynini.closure(punct + pynutil.insert(" "))
+                + token
+                + pynini.closure(pynutil.insert(" ") + punct)
+            )
+
+            graph = token_plus_punct + pynini.closure(
+                pynutil.add_weight(delete_extra_space, 1.1) + token_plus_punct
+            )
+
+            graph = delete_space + graph + delete_space
+            self.fst = graph.optimize()
+
+            if far_file:
+                generator_main(far_file, {"tokenize_and_classify": self.fst})
+                logging.info(f"ClassifyFst grammars are saved to {far_file}.")
@@ -0,0 +1,21 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import GraphFst, convert_space
+from fun_text_processing.text_normalization.ru.utils import get_abs_path
+from pynini.lib import pynutil
+
+
+class WhiteListFst(GraphFst):
+    """
+    Finite state transducer for classifying whitelist, e.g.
+        "квартира" -> telephone { number_part: "кв." }
+
+    Args:
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, deterministic: bool = True):
+        super().__init__(name="whitelist", kind="classify", deterministic=deterministic)
+        whitelist = pynini.string_file(get_abs_path("data/whitelist.tsv")).invert()
+        graph = pynutil.insert('name: "') + convert_space(whitelist) + pynutil.insert('"')
+        self.fst = graph.optimize()
@@ -0,0 +1 @@
+
@@ -0,0 +1,41 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import (
+    DAMO_NOT_QUOTE,
+    GraphFst,
+    delete_space,
+)
+from pynini.lib import pynutil
+
+
+class CardinalFst(GraphFst):
+    """
+    Finite state transducer for verbalizing roman numerals
+        e.g. cardinal { integer: "1 001" } -> 1 001
+
+    Args:
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, deterministic: bool = True):
+        super().__init__(name="cardinal", kind="verbalize", deterministic=deterministic)
+
+        optional_sign = pynini.closure(
+            pynutil.delete("negative:")
+            + delete_space
+            + pynutil.delete('"')
+            + DAMO_NOT_QUOTE
+            + pynutil.delete('"')
+            + delete_space,
+            0,
+            1,
+        )
+
+        graph = (
+            optional_sign
+            + pynutil.delete('integer: "')
+            + pynini.closure(DAMO_NOT_QUOTE, 1)
+            + pynutil.delete('"')
+        )
+        delete_tokens = self.delete_tokens(graph)
+        self.fst = delete_tokens.optimize()
@@ -0,0 +1,16 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import DAMO_NOT_QUOTE, GraphFst
+from pynini.lib import pynutil
+
+
+class DateFst(GraphFst):
+    """
+    Finite state transducer for verbalizing date, e.g.
+        date { day: "02.03.89" }  -> "02.03.89"
+    """
+
+    def __init__(self):
+        super().__init__(name="date", kind="verbalize")
+        graph = pynutil.delete('day: "') + pynini.closure(DAMO_NOT_QUOTE, 1) + pynutil.delete('"')
+        delete_tokens = self.delete_tokens(graph.optimize())
+        self.fst = delete_tokens.optimize()
@@ -0,0 +1,35 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import (
+    DAMO_NOT_QUOTE,
+    DAMO_SPACE,
+    GraphFst,
+    delete_space,
+)
+from pynini.lib import pynutil
+
+
+class DecimalFst(GraphFst):
+    """
+    Finite state transducer for verbalizing decimal, e.g.
+        decimal { negative: "true" integer_part: "3," fractional_part: "2" } -> -3,2
+
+    Args:
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, deterministic: bool = True):
+        super().__init__(name="decimal", kind="verbalize", deterministic=deterministic)
+
+        optional_sign = pynini.closure(pynini.cross('negative: "true" ', "-"), 0, 1)
+
+        integer = pynutil.delete(' "') + pynini.closure(DAMO_NOT_QUOTE, 1) + pynutil.delete('"')
+        integer_part = pynutil.delete("integer_part:") + integer
+        fractional_part = pynutil.delete("fractional_part:") + integer
+        optional_quantity = pynini.closure(
+            pynini.accep(DAMO_SPACE) + pynutil.delete("quantity:") + integer, 0, 1
+        )
+
+        graph = optional_sign + integer_part + delete_space + fractional_part + optional_quantity
+        delete_tokens = self.delete_tokens(graph)
+        self.fst = delete_tokens.optimize()
@@ -0,0 +1,19 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import DAMO_NOT_QUOTE, GraphFst
+from pynini.lib import pynutil
+
+
+class ElectronicFst(GraphFst):
+    """
+    Finite state transducer for verbalizing electronic
+        e.g. electronic { username: "ab@nd.ru" } -> "ab@nd.ru"
+    """
+
+    def __init__(self):
+        super().__init__(name="electronic", kind="verbalize")
+
+        graph = (
+            pynutil.delete('username: "') + pynini.closure(DAMO_NOT_QUOTE, 1) + pynutil.delete('"')
+        )
+        delete_tokens = self.delete_tokens(graph)
+        self.fst = delete_tokens.optimize()
@@ -0,0 +1,27 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import (
+    DAMO_NOT_QUOTE,
+    GraphFst,
+    delete_space,
+)
+from pynini.lib import pynutil
+
+
+class MeasureFst(GraphFst):
+    """
+    Finite state transducer for verbalizing measure
+        e.g. measure { cardinal { integer: "2 кг" } } -> "2 кг"
+    """
+
+    def __init__(self):
+        super().__init__(name="measure", kind="verbalize")
+
+        graph = (
+            pynutil.delete(' cardinal { integer: "')
+            + pynini.closure(DAMO_NOT_QUOTE, 1)
+            + pynutil.delete('"')
+            + delete_space
+            + pynutil.delete("}")
+        )
+        delete_tokens = self.delete_tokens(graph)
+        self.fst = delete_tokens.optimize()
@@ -0,0 +1,21 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import DAMO_NOT_QUOTE, GraphFst
+from pynini.lib import pynutil
+
+
+class MoneyFst(GraphFst):
+    """
+    Finite state transducer for verbalizing electronic
+        e.g. money { integer_part: "2 руб." } -> "2 руб."
+    """
+
+    def __init__(self):
+        super().__init__(name="money", kind="verbalize")
+
+        graph = (
+            pynutil.delete('integer_part: "')
+            + pynini.closure(DAMO_NOT_QUOTE, 1)
+            + pynutil.delete('"')
+        )
+        delete_tokens = self.delete_tokens(graph)
+        self.fst = delete_tokens.optimize()
@@ -0,0 +1,22 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import DAMO_NOT_QUOTE, GraphFst
+from pynini.lib import pynutil
+
+
+class OrdinalFst(GraphFst):
+    """
+    Finite state transducer for verbalizing ordinal numbers
+        e.g. ordinal { integer: "2" } -> "2"
+
+    Args:
+        deterministic: if True will provide a single transduction option,
+            for False multiple transduction are generated (used for audio-based normalization)
+    """
+
+    def __init__(self, deterministic: bool = True):
+        super().__init__(name="ordinal", kind="verbalize", deterministic=deterministic)
+
+        value = pynini.closure(DAMO_NOT_QUOTE)
+        graph = pynutil.delete('integer: "') + value + pynutil.delete('"')
+        delete_tokens = self.delete_tokens(graph)
+        self.fst = delete_tokens.optimize()
@@ -0,0 +1,21 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import DAMO_NOT_QUOTE, GraphFst
+from pynini.lib import pynutil
+
+
+class TelephoneFst(GraphFst):
+    """
+    Finite state transducer for verbalizing telephone
+        e.g. telephone { number_part: "8-913-983-56-01" } -> "8-913-983-56-01"
+    """
+
+    def __init__(self):
+        super().__init__(name="telephone", kind="verbalize")
+
+        graph = (
+            pynutil.delete('number_part: "')
+            + pynini.closure(DAMO_NOT_QUOTE, 1)
+            + pynutil.delete('"')
+        )
+        delete_tokens = self.delete_tokens(graph)
+        self.fst = delete_tokens.optimize()
@@ -0,0 +1,40 @@
+import pynini
+from fun_text_processing.text_normalization.en.graph_utils import (
+    DAMO_NOT_QUOTE,
+    GraphFst,
+    delete_space,
+)
+from pynini.lib import pynutil
+
+
+class TimeFst(GraphFst):
+    """
+    Finite state transducer for verbalizing time
+        e.g. time { hours: "02:15" } -> "02:15"
+    """
+
+    def __init__(self):
+        super().__init__(name="time", kind="verbalize")
+        hour = (
+            pynutil.delete("hours: ")
+            + pynutil.delete('"')
+            + pynini.closure(DAMO_NOT_QUOTE, 1)
+            + pynutil.delete('"')
+        )
+        minutes = (
+            pynutil.delete("minutes: ")
+            + pynutil.delete('"')
+            + pynini.closure(DAMO_NOT_QUOTE, 1)
+            + pynutil.delete('"')
+        )
+
+        graph_preserve_order = (
+            pynutil.delete('hours: "') + pynini.closure(DAMO_NOT_QUOTE, 1) + pynutil.delete('"')
+        )
+
+        # for cases that require permutations for the correct verbalization
+        graph_reverse_order = hour + delete_space + pynutil.insert(":") + minutes + delete_space
+
+        graph = graph_preserve_order | graph_reverse_order
+        delete_tokens = self.delete_tokens(graph)
+        self.fst = delete_tokens.optimize()
@@ -0,0 +1,50 @@
+from fun_text_processing.inverse_text_normalization.en.verbalizers.whitelist import WhiteListFst
+from fun_text_processing.inverse_text_normalization.ru.verbalizers.cardinal import CardinalFst
+from fun_text_processing.inverse_text_normalization.ru.verbalizers.date import DateFst
+from fun_text_processing.inverse_text_normalization.ru.verbalizers.decimal import DecimalFst
+from fun_text_processing.inverse_text_normalization.ru.verbalizers.electronic import ElectronicFst
+from fun_text_processing.inverse_text_normalization.ru.verbalizers.measure import MeasureFst
+from fun_text_processing.inverse_text_normalization.ru.verbalizers.money import MoneyFst
+from fun_text_processing.inverse_text_normalization.ru.verbalizers.ordinal import OrdinalFst
+from fun_text_processing.inverse_text_normalization.ru.verbalizers.telephone import TelephoneFst
+from fun_text_processing.inverse_text_normalization.ru.verbalizers.time import TimeFst
+from fun_text_processing.text_normalization.en.graph_utils import GraphFst
+
+
+class VerbalizeFst(GraphFst):
+    """
+    Composes other verbalizer grammars.
+    For deployment, this grammar will be compiled and exported to OpenFst Finate State Archiv (FAR) File.
+    More details to deployment at NeMo/tools/text_processing_deployment.
+    """
+
+    def __init__(self):
+        super().__init__(name="verbalize", kind="verbalize")
+        cardinal = CardinalFst()
+        cardinal_graph = cardinal.fst
+        ordinal = OrdinalFst()
+        ordinal_graph = ordinal.fst
+        decimal = DecimalFst()
+        decimal_graph = decimal.fst
+        whitelist_graph = WhiteListFst().fst
+        electronic_graph = ElectronicFst().fst
+        money_graph = MoneyFst().fst
+        date_graph = DateFst().fst
+        measure_graph = MeasureFst().fst
+        telephone_graph = TelephoneFst().fst
+        time_graph = TimeFst().fst
+
+        graph = (
+            whitelist_graph
+            | cardinal_graph
+            | ordinal_graph
+            | decimal_graph
+            | electronic_graph
+            | date_graph
+            | money_graph
+            | measure_graph
+            | telephone_graph
+            | time_graph
+        )
+
+        self.fst = graph
@@ -0,0 +1,33 @@
+import pynini
+from fun_text_processing.inverse_text_normalization.en.verbalizers.word import WordFst
+from fun_text_processing.inverse_text_normalization.ru.verbalizers.verbalize import VerbalizeFst
+from fun_text_processing.text_normalization.en.graph_utils import (
+    GraphFst,
+    delete_extra_space,
+    delete_space,
+)
+from pynini.lib import pynutil
+
+
+class VerbalizeFinalFst(GraphFst):
+    """
+    Finite state transducer that verbalizes an entire sentence, e.g.
+    tokens { name: "its" } tokens { time { hours: "12" minutes: "30" } } tokens { name: "now" } -> its 12:30 now
+    """
+
+    def __init__(self):
+        super().__init__(name="verbalize_final", kind="verbalize")
+        verbalize = VerbalizeFst().fst
+        word = WordFst().fst
+        types = verbalize | word
+        graph = (
+            pynutil.delete("tokens")
+            + delete_space
+            + pynutil.delete("{")
+            + delete_space
+            + types
+            + delete_space
+            + pynutil.delete("}")
+        )
+        graph = delete_space + pynini.closure(graph + delete_extra_space) + graph + delete_space
+        self.fst = graph