zhifu gao
2024-05-08 4adb76a6edbca93aae7caa83382e764d7b058f07
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
import os
import string
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
 
FUN_CHAR = utf8.VALID_UTF8_CHAR
 
FUN_DIGIT = byte.DIGIT
FUN_LOWER = pynini.union(*string.ascii_lowercase).optimize()
FUN_UPPER = pynini.union(*string.ascii_uppercase).optimize()
FUN_ALPHA = pynini.union(FUN_LOWER, FUN_UPPER).optimize()
 
FUN_SPACE = " "
FUN_WHITE_SPACE = pynini.union(" ", "\t", "\n", "\r", "\u00A0").optimize()
FUN_NOT_SPACE = pynini.difference(FUN_CHAR, FUN_WHITE_SPACE).optimize()
FUN_NOT_QUOTE = pynini.difference(FUN_CHAR, r'"').optimize()
 
FUN_PUNCT = pynini.union(*map(pynini.escape, string.punctuation)).optimize()
 
 
FUN_SIGMA = pynini.closure(FUN_CHAR)
 
delete_space = pynutil.delete(pynini.closure(FUN_WHITE_SPACE))
delete_zero_or_one_space = pynutil.delete(pynini.closure(FUN_WHITE_SPACE, 0, 1))
insert_space = pynutil.insert(" ")
delete_extra_space = pynini.cross(pynini.closure(FUN_WHITE_SPACE, 1), " ")
 
 
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()
    print(f"Created {file_name}")
 
 
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("\u00A0", " "), "", "", FUN_SIGMA)