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
import pynini
from fun_text_processing.text_normalization.en.graph_utils import (
DAMO_CHAR,
DAMO_DIGIT,
GraphFst,
delete_space,
insert_space,
)
from pynini.lib import pynutil
class TimeFst(GraphFst):
"""
Finite state transducer for verbalizing time,
e.g. time { hours: "la 1" minutes: "10" } -> la 1:10
e.g. time { hours: "la 1" minutes: "45" } -> la 1:45
"""
def __init__(self):
super().__init__(name="time", kind="verbalize")
add_leading_zero_to_double_digit = (DAMO_DIGIT + DAMO_DIGIT) | (
pynutil.insert("0") + DAMO_DIGIT
)
# hour includes preposition ("la" or "las")
hour = (
pynutil.delete("hours:")
+ delete_space
+ pynutil.delete('"')
+ pynini.union("la ", "las ")
+ pynini.closure(DAMO_DIGIT, 1)
+ pynutil.delete('"')
)
minute = (
pynutil.delete("minutes:")
+ delete_space
+ pynutil.delete('"')
+ pynini.closure(DAMO_DIGIT, 1)
+ pynutil.delete('"')
)
suffix = (
delete_space
+ insert_space
+ pynutil.delete("suffix:")
+ delete_space
+ pynutil.delete('"')
+ pynini.closure(DAMO_CHAR - " ", 1)
+ pynutil.delete('"')
)
optional_suffix = pynini.closure(suffix, 0, 1)
zone = (
delete_space
+ insert_space
+ pynutil.delete("zone:")
+ delete_space
+ pynutil.delete('"')
+ pynini.closure(DAMO_CHAR - " ", 1)
+ pynutil.delete('"')
)
optional_zone = pynini.closure(zone, 0, 1)
graph = (
hour
+ delete_space
+ pynutil.insert(":")
+ (minute @ add_leading_zero_to_double_digit)
+ optional_suffix
+ optional_zone
)
delete_tokens = self.delete_tokens(graph)
self.fst = delete_tokens.optimize()