0.7.0 2026-06-08
        - conllu2korapxml:
            - Add automatic offset computation from 'text' comments
              (implemented with AI assistance)
            - new option --text-sigle <template> to derive text sigle from newdoc id
              (implemented with AI assistance)
            - Add --base-text option to generate data.xml with
              reconstructed plain text from 'text' comment lines
              (implemented with AI assistance)
            - Add --base-tokens option to generate base/tokens.xml
              (implemented with AI assistance)
            - Filter SpaceAfter/SpacesAfter from MISC column
              (implemented with AI assistance)
            - Add struct.xml generation from sent_id and newpar comments
              (implemented with AI assistance)

0.6.3 2024-06-04
        - Trim filenames to fix double space after filename metadata
        - Set permissions of zip contents to 666
        - Allow different foundries for morpho and dependency annotations
        - Readme: Drop early stage warning
        - Drop cherry-pick unfriendly test count prediction
        - conllu2korapxml:
            - escape &, <, >
            - convert upos column to upos features

0.6.2 2024-01-24
        - Bump minimal perl version to 5.36 to improve unicode handling.
        - korapxml2conllu
            - Use implicit default utf8 encoding instead of explicit de/encodes. Speeds up processing by 10%.

0.6.1 2023-03-22
        - conllu2korapxml:
            - Fix append for filehandle output.

0.6.0 2023-01-13
        - korapxml2conllu:
            - the sigle-pattern option now affects the entire sigle
            - handle docid attributes correctly if they are in a different line than their parent element <layer>
            - Improve identification of offset errors

0.5.0 2022-09-29
        - korapxml2conllu:
            - --word2vec|lm-training-data option added to print word2vec input format
            - --extract-metadata-regex added to extract some metadata values as context input for language model training
            - by default sentence boundary information is now read from structure.xml files (use --s-bounds-from-morpho otherwise)
            - use morpho.xml if present when run on base zips
            - new option -c <columns>
        - conllu2korapxml:
            - ignore _-lemmas
            - handle UDPipe comments
            - ignore non-interpretable comments
            - improve error handling for missing text ids and offsets

0.4.1 2021-07-31
        - korapxml2conllu: fix patterns not extracted for last texts in archive

0.4 2021-07-29
        - korapxml2conllu option -e <regex> added to extract element/attributes to comments

0.3 2021-02-15
        - Provide conllu2korapxml to convert from ConLL-U to KorAP-XML zip

0.2 2021-02-12
        - Convert also KorAP-XML base zips

0.1 2020-09-23
        - Initial release to GitHub.