Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
Python asraudiochineseemotion-recognitionfunasrmcp-servermultilingual-asropenai-compatible-apiparaformerpunctuationpytorchreal-time-asrspeaker-diarizationspeech-recognitionspeech-to-textstreaming-asrtranscriptionvllmvoice-activity-detectionwhisper-alternative
from fun_text_processing.inverse_text_normalization.zh.graph_utils import (
DAMO_NOT_QUOTE,
GraphFst,
delete_extra_space,
delete_space,
)
import pynini
from pynini.lib import pynutil
class FractionFst(GraphFst):
"""
Finite state transducer for verbalizing fraction,
"""
def __init__(self):
super().__init__(name="fraction", kind="verbalize")
optional_sign = pynini.closure(pynini.cross('negative: "true"', "-") + delete_space, 0, 1)
numerator = (
pynutil.delete("numerator:")
+ delete_space
+ pynutil.delete('"')
+ pynini.closure(DAMO_NOT_QUOTE, 1)
+ pynutil.delete('"')
)
denominator = (
pynutil.delete("denominator:")
+ delete_space
+ pynutil.delete('"')
+ pynini.closure(DAMO_NOT_QUOTE, 1)
+ pynutil.delete('"')
)
graph = (
optional_sign + numerator + delete_space + pynutil.insert("/") + denominator
).optimize()
self.numbers = graph
delete_tokens = self.delete_tokens(optional_sign + graph)
self.fst = delete_tokens.optimize()