提交

kongdeqiang

2026-03-13 28ccfbfc51068a663a80764e14074df5edf2b5ba

 fun_text_processing/inverse_text_normalization/fr/verbalizers/ordinal.py

@@ -1,17 +1,3 @@
# Copyright NeMo (https://github.com/NVIDIA/NeMo). 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.

import pynini
from fun_text_processing.inverse_text_normalization.fr.graph_utils import (
    DAMO_DIGIT,
@@ -39,9 +25,9 @@
        graph_integer = (
            pynutil.delete("integer:")
            + delete_space
            + pynutil.delete("\"")
            + pynutil.delete('"')
            + pynini.closure(DAMO_NOT_QUOTE, 1)
            + pynutil.delete("\"")
            + pynutil.delete('"')
        )

        replace_suffix = pynini.union(
@@ -50,22 +36,26 @@
            pynini.cross("r", "ʳ"),
            pynini.cross("s", "ˢ"),
        )
        replace_suffix = pynutil.delete(" morphosyntactic_features: \"") + replace_suffix.plus
        replace_suffix = pynutil.delete(' morphosyntactic_features: "') + replace_suffix.plus

        graph_arabic = graph_integer + replace_suffix.plus

        # For roman.
        graph_roman_digits = pynini.string_file(get_abs_path("data/roman/digits_large.tsv")).invert()
        graph_roman_digits = pynini.string_file(
            get_abs_path("data/roman/digits_large.tsv")
        ).invert()
        graph_roman_ties = pynini.string_file(get_abs_path("data/roman/ties_large.tsv")).invert()
        graph_roman_hundreds = pynini.string_file(get_abs_path("data/roman/hundreds_large.tsv")).invert()
        graph_roman_hundreds = pynini.string_file(
            get_abs_path("data/roman/hundreds_large.tsv")
        ).invert()
        graph_roman_zero_digit = pynutil.delete("0")

        graph_roman_hundreds = DAMO_DIGIT ** 3 @ (
        graph_roman_hundreds = DAMO_DIGIT**3 @ (
            graph_roman_hundreds
            + pynini.union(graph_roman_ties, graph_roman_zero_digit)
            + pynini.union(graph_roman_digits, graph_roman_zero_digit)
        )
        graph_roman_ties = DAMO_DIGIT ** 2 @ (
        graph_roman_ties = DAMO_DIGIT**2 @ (
            graph_roman_ties + pynini.union(graph_roman_digits, graph_roman_zero_digit)
        )
        graph_roman_digits = DAMO_DIGIT @ graph_roman_digits
@@ -75,7 +65,7 @@
        graph_roman = (graph_integer @ graph_roman_integers) + replace_suffix
        graph_roman += pynini.cross("/", " ") + "siècle"

        graph = (graph_roman | graph_arabic) + pynutil.delete("\"")
        graph = (graph_roman | graph_arabic) + pynutil.delete('"')

        delete_tokens = self.delete_tokens(graph)
        self.fst = delete_tokens.optimize()