timenorm-1.0.5

• fullSpan is added when created TimeZone in Readers

timenorm-1.0.4

This release includes some fixes:

• The regex to clean the source text captures unicode-32 characters.
• The neural parser removes repeating operators for same span.
• For IntersectionRI: peel off correctly intervals that are before startPoint.

timenorm-1.0.3

This release includes some improvements in the linking process in the scala version of the parser, especially for the Between operator:

• linking of the Between operator has been improved checking the text and the position with respect to the child entity. E.g. "until 2019" cannot be linked with End-Interval.
• Unknown is now the default value for Type property if the case is not found in date-types.txt
• For Frequency the value of the Type operator is Other.
• date-types.txt has been refined removing unnecessary cases and adding new ones like hours, seconds,...

In parseBatchToXML the text is cleaned replacing one control characters with one space character to avoid wrong spans when the text contains consecutive control characters (^C, ^L, ...). Previously, consecutive control characters were replaced with a single space character.

timenorm-1.0.2

• Throws an early exception when trying to normalize impossible dates, e.g. "April 31".
• Set isDefined = false for Between operator if startInterval ends after endInterval begins.

timenorm-1.0.1

v1.0.1

Handles cases where non-numbers (unparseable inputs) are passed to Wo…

timenorm-1.0.0

This release includes a number of major changes:

• Packages have been reorganized to separate the new time normalizer based on the SCATE schema and the old time normalizer based on the TimeML schema.
• The neural SCATE-based normalizer now directly returns the SCATE operators (TimeExpression objects), which now include character offsets of where each operator was found.
• The anti-xml dependency has been removed, using the standard scala-xml instead. The scala-xml APIs are ugly, but anti-xml is no longer maintained.
• The English number parsing grammar that was included as part of the old time normalizer has been simplified and separated out into its own API, WordsToNumber, so it can be used by the neural parser.
• The build is now based on SBT instead of Maven.

timenorm-0.12.1

Complete backwards the annotation starting from (start - 1) character…

timenorm-0.12.0

Adds analysis of distance between linked entities, broken down by pro…

timenorm-0.11.2

The entities can be passed to the parse method as an option so the id…

timenorm-0.11.1_2.11.11

Minor