Hungarian Date Parser

A tool for extracting datetime intervals from Hungarian sentences and turning datetime objects into Hungarian text.

Install and try the package with pip install hun-date-parser

🔥 Usage

from hun_date_parser import text2datetime
from datetime import datetime

text2datetime('találkozzunk jövő kedd délután!', now=datetime(2020, 12, 27))
# [{'start_date': datetime.datetime(2020, 12, 29, 12, 0), 'end_date': datetime.datetime(2020, 12, 29, 17, 59, 59)}]

text2datetime('találkozzunk jövő héten szombaton háromnegyed nyolc előtt két perccel', now=datetime(2020, 12, 27))
# [{'start_date': datetime.datetime(2021, 1, 2, 7, 43), 'end_date': datetime.datetime(2021, 1, 2, 7, 43, 59)}]

text2datetime('találkozzunk jövő héten szombaton este háromnegyed nyolc előtt két perccel', now=datetime(2020, 12, 27))
# [{'start_date': datetime.datetime(2021, 1, 2, 19, 43), 'end_date': datetime.datetime(2021, 1, 2, 19, 43, 59)}]

The date parser is also capable of parsing explicit intervals from the text even when only one side of the interval is specified.

from hun_date_parser import text2datetime
from datetime import datetime

text2datetime('2020 decemberétől 2021 januárig', now=datetime(2020, 12, 27))
# [{'start_date': datetime.datetime(2020, 12, 1, 0, 0), 'end_date': datetime.datetime(2021, 1, 31, 23, 59, 59)}]

text2datetime('2021 januárig', now=datetime(2020, 12, 27))
# [{'start_date': None, 'end_date': datetime.datetime(2021, 1, 31, 23, 59, 59)}]

If not specified otherwise, relative dates (eg.: tomorrow, next week, etc.) are calculated relative to the current datetime, at the time when the function is called. The now parameter can be used for parsing relative datetimes relative to any timestamp other than the current time.

Supported formats

Our parser implements a rule-based strategy to interpret a diverse array of date and time formats, utilizing grammatical inflection to parse intervals.

The following formats are currently supported:

Date Formats:

• ISO Standard Dates: dates formatted to the ISO 8601 standard.
  • Examples: 2020-01-15, 2020-12-30-án, 2020.12.29.
• Named Months: months indicated by name, optionally with day numbers and/or year. Day numbers may be expressed lexically.
  • Examples: tavaly február, 2020 február 3, jövő március, jövő február 12-én, március elsején.
• Relative Time References: relative days, weeks, months, or years.
  • Examples: tegnap, ma, holnap, idén, tavaly, múlt héten, a múlt hónapban, idei események.
• Named Days of the Week: references to specific weekdays, accounting for past, present, and future contexts.
  • Examples: múlt vasárnap, kedden, ezen a heten hétfőn, jövő héten szerdán.
• Counted Time Frames: expressions indicating a number of days or weeks ago or in the future.
  • Examples: 1 héttel ezelőtt, 6 nappal ezelőtt, 5 nap múlva.
• Historical Periods: periods up to the present date, defined by days, weeks, months, or years.
  • Examples: az előző két hétben, az előző két évi adatok, az előző 10 nap eredménye.

Time Formats:

• Digital Clock Format: time expressed in digital clock notation.
  • Examples: 18:12-kor, 06:45.
• Natural Language Time: time described in conversational terms.
  • Examples: este fél 8, reggel nyolc előtt hat perccel, nyolc óra nyolc perckor, 20 óra 49 perckor, este negyed 8 előtt 6 perccel.
• Abbreviated Time Formats: certain abbreviated forms of time expression.
  • Examples: 16h-kor.

Interval Formats:

• Inflection-Implied Ranges: intervals detected through grammatical inflections, indicating a start and end point.
  • Examples: február 13-tól 17-ig, keddtől péntekig, januártól februárig, 2020-10-12-től 2020-11-01-ig.
• Open-Ended Intervals: expressions where the start or end of the interval is unspecified, using inflections to imply boundaries.
  • Examples: keddtől, február elsejéig.

Setting search scope in case of ambiguous input

For the function text2datetime, the parameter search_scope is used to specify the desired time interval for parsing inputs.

• The default value, SearchScopes.NOT_RESTRICTED, does not restrict whether the scope of the search is in the past or the future.
  • For example, when Tuesday is parsed, the date for the Tuesday of the given week will be returned, without considering whether that date is in the past or the future.
• To prefer future dates in case of ambiguity, use the value SearchScopes.FUTURE_DAY.
  • In this case, when Tuesday is parsed, the function will return the nearest Tuesday in the future, not necessarily the current week's Tuesday.
• Similarly, to search in the past, nudging the library to prefer past dates is possible with the value SearchScopes.PAST_SEARCH.
  • For instance, if May is parsed by the function, with this setting, and if this year's May is still in the future, last year's May will be returned.
  • Please note that when there's no ambiguity, the function can still return future or past dates, even when a different preference is specified.

The flag realistic_year_required can be set in order to minimize false matches, it generally restricts year mentions to be between 1900 and 2100.

• It defaults to true.
• As the number 3000 is unlikely to be considered a year value in everyday contexts, it is ignored.
• The output datetime can still be later than 2100 or earlier than 1900 with mentions of 'Counted Time Frames' (e.g., 100000 nap múlva).

An example:

from hun_date_parser import text2datetime
from datetime import datetime
from hun_date_parser.utils import SearchScopes

text2datetime('augusztus', now=datetime(2023, 6, 7), search_scope=SearchScopes.PAST_SEARCH)
# [{'start_date': datetime.datetime(2022, 8, 1, 0, 0),
#   'end_date': datetime.datetime(2022, 8, 31, 23, 59, 59)}]

text2datetime('péntek', now=datetime(2023, 6, 7), search_scope=SearchScopes.PAST_SEARCH)
# [{'start_date': datetime.datetime(2023, 6, 2, 0, 0),
#   'end_date': datetime.datetime(2023, 6, 2, 23, 59, 59)}]

text2datetime('péntek', now=datetime(2023, 6, 7), search_scope=SearchScopes.NOT_RESTRICTED)
# [{'start_date': datetime.datetime(2023, 6, 9, 0, 0),
#   'end_date': datetime.datetime(2023, 6, 9, 23, 59, 59)}]

Duration Parsing

The duration parser can extract the duration in minutes from various expressions found in sentences.

Recognized Formats

The parser is capable of understanding a variety of duration expressions. Here are the primary formats it recognizes:

• Hour and Minute Combination:
  • Examples: 1 óra 45 perc, egy óra 30 perc, 2 óra 15 perc
• Hour Only:
  • Examples: 1 óra, egy óra, 2 órát, 3,5 óra
• Quarter Hour Phrases:
  • Examples: háromnegyed óra, egy és negyed óra, kettő és fél óra

from hun_date_parser import parse_duration

print(parse_duration('45 perc'))  # Output: 45

print(parse_duration('1 és negyed óra'))  # Output: 75

Frequency Parsing

The frequency parser can identify recurring time patterns from Hungarian text expressions and returns the match position information.

Recognized Formats

The parser recognizes standard frequency expressions in Hungarian:

• Daily Frequencies:
  • Examples: napi, naponta, minden nap
• Weekly Frequencies:
  • Examples: heti, hetente, minden héten, heti rendszerességgel
• Fortnightly Frequencies:
  • Examples: kétheti, kéthetente
• Monthly Frequencies:
  • Examples: havi, havonta, minden hónapban, havi rendszerességgel
• Quarterly Frequencies:
  • Examples: negyedévente, minden negyedévben
• Half-yearly Frequencies:
  • Examples: félévente, minden félévben
• Yearly Frequencies:
  • Examples: évente, minden évben

from hun_date_parser import parse_frequency
from hun_date_parser.frequency_parser import Frequency

# The function returns a dictionary with frequency value and position information
result = parse_frequency('napi')
# result: {'frequency': Frequency.DAILY, 'start': 0, 'end': 4}

print(parse_frequency('hetente')['frequency'])  # Output: Frequency.WEEKLY
print(parse_frequency('kétheti')['frequency'])  # Output: Frequency.FORTNIGHTLY
print(parse_frequency('havonta')['frequency'])  # Output: Frequency.MONTHLY
print(parse_frequency('negyedévente')['frequency'])  # Output: Frequency.QUARTERLY
print(parse_frequency('félévente')['frequency'])  # Output: Frequency.EVERY_HALF_YEAR
print(parse_frequency('évente')['frequency'])  # Output: Frequency.YEARLY

# Since Frequency is a string-based enum, values can be used as strings
print(str(parse_frequency('naponta')['frequency']))  # Output: "DAILY"

# Example with text context and positions
result = parse_frequency('Találkozzunk hetente a parkban')
# result: {'frequency': Frequency.WEEKLY, 'start': 12, 'end': 19}

Current Limitations

• The parser doesn't yet support complex frequencies like naponta kétszer (twice a day)
• Specific weekday expressions like minden kedden (every Tuesday) aren't currently recognized
• Expressions like minden második pénteken (every second Friday) or minden hónap első hétfőjén (first Monday of every month) aren't supported

Datetime to text

The library is also capable of turning datetime objects into their Hungarian text representation.

from hun_date_parser import datetime2text
from datetime import datetime

datetime2text(datetime(2020, 12, 20, 18, 34), now=datetime(2020, 12, 27), time_precision=2)
# {'dates': ['múlt héten vasárnap', '2020 december 20'],
#  'times': ['tizennyolc óra harmincnégy perc', '18:34', 'este hat óra harmincnégy perc', 'este fél 7 után 4 perccel']}

📝 License

This project is licensed under MIT license. Feel free to use it in your own projects.

🔧 Contribute

Any help or feedback in further developing the library is welcome!