electronic.py 2.9 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566
  1. import pynini
  2. from fun_text_processing.text_normalization.en.graph_utils import (
  3. DAMO_NOT_QUOTE,
  4. DAMO_SIGMA,
  5. GraphFst,
  6. delete_preserve_order,
  7. insert_space,
  8. )
  9. from fun_text_processing.text_normalization.es.utils import get_abs_path
  10. from pynini.lib import pynutil
  11. digit_no_zero = pynini.invert(pynini.string_file(get_abs_path("data/numbers/digit.tsv")))
  12. zero = pynini.invert(pynini.string_file(get_abs_path("data/numbers/zero.tsv")))
  13. graph_symbols = pynini.string_file(get_abs_path("data/electronic/symbols.tsv"))
  14. server_common = pynini.string_file(get_abs_path("data/electronic/server_name.tsv"))
  15. domain_common = pynini.string_file(get_abs_path("data/electronic/domain.tsv"))
  16. class ElectronicFst(GraphFst):
  17. """
  18. Finite state transducer for verbalizing electronic
  19. e.g. electronic { username: "abc" domain: "hotmail.com" } -> "a b c arroba hotmail punto com"
  20. -> "a b c arroba h o t m a i l punto c o m"
  21. -> "a b c arroba hotmail punto c o m"
  22. -> "a b c at h o t m a i l punto com"
  23. Args:
  24. deterministic: if True will provide a single transduction option,
  25. for False multiple transduction are generated (used for audio-based normalization)
  26. """
  27. def __init__(self, deterministic: bool = True):
  28. super().__init__(name="electronic", kind="verbalize", deterministic=deterministic)
  29. graph_digit_no_zero = (
  30. digit_no_zero @ pynini.cdrewrite(pynini.cross("un", "uno"), "", "", DAMO_SIGMA).optimize()
  31. )
  32. graph_digit = graph_digit_no_zero | zero
  33. def add_space_after_char():
  34. return pynini.closure(DAMO_NOT_QUOTE - pynini.accep(" ") + insert_space) + (
  35. DAMO_NOT_QUOTE - pynini.accep(" ")
  36. )
  37. verbalize_characters = pynini.cdrewrite(graph_symbols | graph_digit, "", "", DAMO_SIGMA)
  38. user_name = pynutil.delete("username: \"") + add_space_after_char() + pynutil.delete("\"")
  39. user_name @= verbalize_characters
  40. convert_defaults = pynutil.add_weight(DAMO_NOT_QUOTE, weight=0.0001) | domain_common | server_common
  41. domain = convert_defaults + pynini.closure(insert_space + convert_defaults)
  42. domain @= verbalize_characters
  43. domain = pynutil.delete("domain: \"") + domain + pynutil.delete("\"")
  44. protocol = (
  45. pynutil.delete("protocol: \"")
  46. + add_space_after_char() @ pynini.cdrewrite(graph_symbols, "", "", DAMO_SIGMA)
  47. + pynutil.delete("\"")
  48. )
  49. self.graph = (pynini.closure(protocol + pynini.accep(" "), 0, 1) + domain) | (
  50. user_name + pynini.accep(" ") + pynutil.insert("arroba ") + domain
  51. )
  52. delete_tokens = self.delete_tokens(self.graph + delete_preserve_order)
  53. self.fst = delete_tokens.optimize()