electronic.py 3.0 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485
  1. import pynini
  2. from fun_text_processing.text_normalization.en.graph_utils import (
  3. DAMO_NOT_QUOTE,
  4. DAMO_NOT_SPACE,
  5. DAMO_SIGMA,
  6. TO_UPPER,
  7. GraphFst,
  8. delete_extra_space,
  9. delete_space,
  10. insert_space,
  11. )
  12. from fun_text_processing.text_normalization.en.utils import get_abs_path
  13. from pynini.examples import plurals
  14. from pynini.lib import pynutil
  15. class ElectronicFst(GraphFst):
  16. """
  17. Finite state transducer for verbalizing electronic
  18. e.g. tokens { electronic { username: "cdf1" domain: "abc.edu" } } -> c d f one at a b c dot e d u
  19. Args:
  20. deterministic: if True will provide a single transduction option,
  21. for False multiple transduction are generated (used for audio-based normalization)
  22. """
  23. def __init__(self, deterministic: bool = True):
  24. super().__init__(name="electronic", kind="verbalize", deterministic=deterministic)
  25. graph_digit_no_zero = pynini.invert(pynini.string_file(get_abs_path("data/number/digit.tsv"))).optimize()
  26. graph_zero = pynini.cross("0", "zero")
  27. if not deterministic:
  28. graph_zero |= pynini.cross("0", "o") | pynini.cross("0", "oh")
  29. graph_digit = graph_digit_no_zero | graph_zero
  30. graph_symbols = pynini.string_file(get_abs_path("data/electronic/symbol.tsv")).optimize()
  31. default_chars_symbols = pynini.cdrewrite(
  32. pynutil.insert(" ") + (graph_symbols | graph_digit) + pynutil.insert(" "), "", "", DAMO_SIGMA
  33. )
  34. default_chars_symbols = pynini.compose(
  35. pynini.closure(DAMO_NOT_SPACE), default_chars_symbols.optimize()
  36. ).optimize()
  37. user_name = (
  38. pynutil.delete("username:")
  39. + delete_space
  40. + pynutil.delete("\"")
  41. + default_chars_symbols
  42. + pynutil.delete("\"")
  43. )
  44. domain_common = pynini.string_file(get_abs_path("data/electronic/domain.tsv"))
  45. domain = (
  46. default_chars_symbols
  47. + insert_space
  48. + plurals._priority_union(
  49. domain_common, pynutil.add_weight(pynini.cross(".", "dot"), weight=0.0001), DAMO_SIGMA
  50. )
  51. + pynini.closure(
  52. insert_space + (pynini.cdrewrite(TO_UPPER, "", "", DAMO_SIGMA) @ default_chars_symbols), 0, 1
  53. )
  54. )
  55. domain = (
  56. pynutil.delete("domain:")
  57. + delete_space
  58. + pynutil.delete("\"")
  59. + domain
  60. + delete_space
  61. + pynutil.delete("\"")
  62. ).optimize()
  63. protocol = pynutil.delete("protocol: \"") + pynini.closure(DAMO_NOT_QUOTE, 1) + pynutil.delete("\"")
  64. graph = (
  65. pynini.closure(protocol + delete_space, 0, 1)
  66. + pynini.closure(user_name + delete_space + pynutil.insert(" at ") + delete_space, 0, 1)
  67. + domain
  68. + delete_space
  69. ).optimize() @ pynini.cdrewrite(delete_extra_space, "", "", DAMO_SIGMA)
  70. delete_tokens = self.delete_tokens(graph)
  71. self.fst = delete_tokens.optimize()