Recognizers-Text

New release of the Recognizers-Text packages to NPM (https://www.npmjs.com/~recognizers.text). Version 1.1.2

Changes

• Multiple bug fixes and refinements to Number and DateTime parsing;
• Multiple fixes for consistency with recognizers in .NET;
• Recognized currencies now include their ISO codes;
• Multiple improvements and extensions to recognition of relative DateTime entities;
• Support for Number, Age, and Currency recognition in Japanese;
• Extensions to French recurring Date/Time Set extraction;
• Improved Number and Time recognition in French;
• Boolean Choice recognizer now supports Chinese and Spanish;
• Improved support for SI units in Chinese;
• Improved Currency recognizer in Chinese;
• Multiple improvements to Sequence recognizer for phone numbers, e-mail addresses, URLs, @mentions;
• New @microsoft/recognizers-text-data-types-timex-expression (no longer under @preview).

Note: Packages unlisted due to regression. Please use release v1.0.8.1

Note: Released packages unlisted due to dependency issue. Please use v1.1.2

Note: Released packages unlisted due to dependency issue. Please use v1.1.2

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.1.0

Changes

• Fixes in Chinese Unit entity boundaries inconsistencies;
• Bug fix in parsing numbers over 70 in the French system;
• Support for French number forms used in Switzerland and Belgium;
• Cardinal numbers recognizer in Dutch;
• Age, Temperature, and Dimensions entity recognizers in Dutch;
• Extracted PhoneNumber entities are now scored based on formatting and length (scoring to be further refined);
• Support for extraction and resolution of “business days” in English DateTime;
• Support for tagging other business terms in DateTime (e.g. “business hours”, “EOD”);
• Support for “decade” as unit of 10Y in English;
• Improved support for relative DateTimes in Chinese;
• Bug fix in handling “upcoming” in relation to DateTime.

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.11

Changes

• Bug fix in Number extraction in Japanese when common written form omits characters;
• Fixes in Chinese date and date period extraction and resolution;
• Cardinal numbers recognizer in Korean;
• Support for week mentions in the form “week of the 22nd”;
• Extended support for typical PhoneNumber formats in the Netherlands, Italy, and Denmark;
• Fix for inconsistent number range extraction in English;
• Improved handling of false positives in time of day mentions in greetings utterances;
• Improved handling of false positives in cases where “one” should not be treated as number;
• Bug fix in handling Time ranges combined with timezones (Preview);
• Boolean choices recognizer in Chinese;
• Improved handling of approximate modifier in DateTime;
• Support "as early/late as" as a modifier for Datetime entities;
• Bug fix in Korean fraction parser.

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.10

Changes

• Fixes in Chinese date and date period extraction and resolution;
• Support for “following” as a relative term applied to datetime;
• Support for more informal ways to specify an amount of day/month/… (e.g. “next couple months”)
• Fix for inconsistent datetime period extraction in English;
• Improved combined recognition of dates and weekdays;
• Improved combined recognition of DayOfWeek + PartOfDay;
• Boolean choices recognizer in Dutch;
• Extensions to Boolean choices recognizer in English;
• Bug fix in Japanese number parser;
• Improvements to datetime extraction with timezones in English (Preview).

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.9

Changes

• Added sub-type information for numbers (e.g. integer, fraction, …), as consumers can leverage this information in post-processing;
• Extensions to Currency recognition in Japanese;
• Fixed inconsistency in dealing with date after a time period;
• Fix in DateTime extraction for ambiguous cases in Chinese;
• Extended recognition of informal mentions of years;
• Improved support for Number recognition as mixed written words and numerals in English;
• Bug fix in extracting multiple durations from the same utterance;
• Performance improvements in DateTimePeriod and base Date extraction;
• Fix encoding issues caused by UTF8+BOM;
• Added support for hyphenated “mid” prefix (e.g. “mid-November”);
• Bug fixes for timezone corner cases in English (Preview).

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.8.2

Changes

• Fixes to month extraction in Spanish/Portuguese;
• Improved entity merge support for date + before/after time cases in English;
• Bug fixes to entity text boundaries and validation;
• Bug fix for regression where a list of two years was treated as a range in English;
• Support for informal time mentions in datetimes like “today at 9a”;
• Bug fixes for datetime range corner cases in English and Chinese.

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.8.1

Changes

• Extended support for SI unit symbols in Chinese;
• Fixes to handle disjoint date ranges in English and Chinese;
• Improved support for different dash characters in ranges (U+2012 to U+2015, U+2053) ;
• Extensions in Japanese Boolean recognizer and bug fixes in Numbers;
• Bug fixes in handling combined less formal ranges + date (e.g. "between 10 and 11:30 on 1/1/2015");
• Support for extracting centuries (e.g. “21st century”) date ranges in English;
• Improved support for “more than/less than” time periods;
• Improved performance on TimeZone matcher and fixes in handling DateTime periods with timezone (Preview);
• Improved URL and Mention extractors;
• Additional holiday support in English;
• Improved support for merged DateTimeAlternatives in list + relative_week cases (e.g. “next week on tuesday or thursday” );
• Enhanced extraction of relative periods from a time point;
• Fix regression where “by year” was interpreted as current year in English;
• Fix for time wrongly tagged within zipcodes;
• Fix "DayOfWeek-Month-Date" recognized as DatePeriod;
• Refinements to year-quarter extraction;
• Fixed resolution for after-year pattern to use correct starting timepoint;
• Bug fixes for number ranges and relative dates in English and Chinese.

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.7

Changes

• Fixes for percent and negative number extraction in Chinese;
• Enhanced extraction of relative dates from a time point;
• Improved handling of later/earlier applied to datetime ranges;
• Additional support for late/early as modifiers for datetime;
• Merged results for datetime + timezone (Preview);
• Preliminary support for “-time timezone” pattern in English markets only;
• Performance improvements to timezone extraction;
• Minor fixes for holiday extraction and informal formats for date + time;
• Bug fixes for date and number ranges.

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.6

Changes

• Improves merging of compound units (currency);
• Fixes for percent extraction in Chinese;
• Bug fixes for year extraction in Chinese;
• Improved phone number recognition;
• Fix ordinal dates not recognized in French;
• Improved support for daterange expressions in English and Chinese;
• Fixes to datetime range resolution when crossing day boundaries;
• Support for same/that + date expressions to differentiate from resolvable dates;
• New ChoiceRecognizer – Boolean in Dutch;
• Improved support for DateTimeAlternatives;
• Improved support for relative terms in English (early, later, etc.);
• Improved handling of ambiguous terms in CalendarMode.

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.5

Changes

• Fix to a longstanding issue with fallback to default language format when parsing ambiguous DateTime;
• Support for additional time formats (HHhMM);
• Improved handling of not well formatted month + year datetimes;
• Bug fixes and extensions to handling multiple DatePeriod in English;
• Fixes to incorrect extraction of fractions and ranges from phone numbers;
• NumberRange, Currency, and Age support in Japanese;
• Fix to regression for simple types in handling alternative datetimes;
• Support for combined Arabic numerals and Chinese counting words;
• Fixes to dot-syntax date support (e.g. 13.5.2015);
• More consistent handling of 2-digit years in date parser;
• Handle "more than" and "less than" modifiers in DateTime;
• Improved support for relative week days in English;
• Extend recognition of DateRange across date types.

New release of the Recognizers-Text packages to NPM (https://www.npmjs.com/~recognizers.text). Version 1.0.1

Changes

• Improvements to false positives in PhoneNumber recognition;
• Improved handling of number ranges;
• Bug fixes and extensions to handling Number and NumberRange in Chinese;
• Fixes to incorrect handling of percent in URLs and to URL extraction;
• Bug fixes for ranges with null resolution in English;
• Support for additional Date formats (e.g. YYYY MM);
• Extensions and bug fixes to relative dates in English;
• Support for number format in standards ISO 31-0 and DIN 1333;
• Improved support for imprecise datetimes;
• Sequence recognizer coverage for e-mail addresses, URLs, @mentions, and #hashtags;
• Multiple extensions to datetime range recognition in English;
• Fixes to IPv6 recognition in unusual cases allowed by standard;
• ChoiceRecognizer now supports Portuguese and Japanese;
• Improved handling of year quarters/halves in Q/H format;
• Better support for negative numbers in English and Chinese;
• New @microsoft/recognizers-text-data-types-timex-expression; which provides parsing and evaluation of TIMEX expressions. Available under @preview.

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.4

Changes

• Ordinal/Percentage/Cardinal number recognition for Japanese;
• Improvements to false positives in PhoneNumber recognition;
• Improved handling of number ranges;
• Bug fixes and extensions to handling Number and NumberRange in Chinese;
• Support for basic lists of alternative DateTime;
• Fixes to incorrect handling of percent in URLs and to URL extraction;
• Extensions to TimeZone support in “preview”;
• Bug fixes for ranges with null resolution in English;
• Support for additional Date formats (e.g. YYYY MM);
• Extensions and bug fixes to relative dates in English;
• Support for number format in standards ISO 31-0 and DIN 1333.

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.3

Changes

• Improved handling of negative numbers and number ranges;
• Bug fixes and extensions to handling multiple DatePeriod in English;
• Fixes to handling fractions and percent;
• Basic Timezone support in “preview”;
• Support for compound currency entities as one entity (“1 dollar and 50 cent” -> “1.5 dollar”);
• Recognized currencies now include their ISO codes (if defined);
• Sequence recognizer coverage for e-mail addresses, URLs, @mentions, and #hashtags;
• Extensions and bug fixes to datetime ranges in English.

This release also includes a new package:
Microsoft.Recognizers.Text.DataTypes.TimexExpression (v1.0.2); which provides parsing and evaluation of TIMEX expressions.

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.2

Changes

• Improved handling of false positive cases for ‘May’ recognized as month;
• Improved handling of DateTime ranges in English;
• Support for “period or next” where the later alternative is implicit (w/ ExtendedTypes);
• Resolution fixes for relative dates in relative ranges (e.g. “second Friday of next month”);
• ChoiceRecognizer now supports Portuguese, Japanese, and Spanish;
• Improved handling of year quarters/halves in Q/H format.

New release of the Recognizers-Text packages to nuget.org (https://www.nuget.org/profiles/Recognizers.Text). Version 1.0.1

Changes

• Recognizer API signature changes:
  • Recognizer initialization is now uniform across packages;
  • Users can either create a specific recognizer instance (according to configuration) or call static Recognize methods for direct access to results (if only default behaviour is needed);
  • Recognizers do not impose singleton behaviour, but cache initialized models for performance.
• New ChoiceRecognizer to extract Boolean entities;
• New IpAddress model in SequenceRecognizer;
• Improved handling of negative numbers in Chinese;
• Extensions and bug fixes to datetime ranges in English.

This is the first public release of the Recognizers-Text packages to NPM (https://www.npmjs.com/package/@microsoft/recognizers-text-suite). The @preview tag is no longer required to install them.

Release Notes

• v1 release of packages:
  • @microsoft/recognizers-text,
  • @microsoft/recognizers-text-number,
  • @microsoft/recognizers-text-number-with-unit,
  • @microsoft/recognizers-text-date-time,
  • @microsoft/recognizers-text-sequence,
  • @microsoft/recognizers-text-choice, and
  • @microsoft/recognizers-text-suite