en_us_normalization.production.classify.TimeFst

class en_us_normalization.production.classify.TimeFst[source]

Finite state transducer for classifying time. Slots to fill when time is parsed:

• hours - 1-2 digit number

• minutes - 1-2 digit number (optional)

• suffix - abbreviation such as AM or PM (optional)

• time zone - abbreviation at the end such as EST (optional)

Hours and minutes can be separated with “:” or “.”. At least one successor should (minutes, suffix or time zone) should help to classify digits as hours.

Examples of input/output strings:

• 12:30 a.m. est -> time { hours: “12” minutes: “30” suffix: “AM” zone: “EST” }

• 2.30 a.m. -> time { hours: “2” minutes: “30” suffix: “AM” }

• 02.30 a.m. -> time { hours: “2” minutes: “30” suffix: “AM” }

• 2.00 a.m. -> time { hours: “2” suffix: “AM” }

• 2 a.m. -> time { hours: “2” suffix: “AM” }

• 02:00 -> time { hours: “2” }

• 2:00 -> time { hours: “2” }

__init__()[source]

static load_shortenings(name: str) → pynini.FstLike[source]

helper function to load time shortenings - suffixes and time zones. produces transducer that allows lower case, upper case, dots-between-letters shortenings from the list