Releases: allenai/mmda
Releases · allenai/mmda
v0.5.0
0.4.8
0.3.0
-
Added
Recipesas a way to combine multiplePredictors& test how they stitch together. Currently added aCoreRecipe. Docs & Tests demonstrate how it can be used. -
Added
Grobidintegration as a way of augmenting an existingDocumentfrom aParser(e.g. PDFPlumber) withGrobidannotations. -
Laying groundwork for a future dataclass called
Relation.
0.2.82
0.2.7
0.2.4
- Added a
Metadataas a type that can exist at a Document-level - Added utility for obtaining
OutlineMetadatafrom a PDF - Fixes to
citation_linkerbecause ofsklearndeprecation - Add
WhiteSpaceTokenizer - Fixes to
DictionaryWordPredictorbecause of change to how tokenization happens inPDFPlumberParser - Change how fieldnames are defined in
types.names - Move off
setup.pyintopyproject.toml
0.1.0
- Changes to Annotation class to remove uuid, require id, change Metadata default behavior
- Changes to JSON serialization schema for Box
- Bugfix in MentionDetector that was changing Document.tokens accidentally due to lack of deepcopy
- Add new predictor for Table/Figure Captions
- Hotfix in PDFPlumberParser that avoids injection of new whitespace in Document.symbols
0.0.44
Add attributes to API data classes (#150) * redesigned apis to account for metadata * bumped version * switched to attributes * simplified code with extra=Extra.ignore explicitly removing `id`, `text`, and `type` is no longer reuqired bc they are automatically ignored. * bumped version, suggestion from @cmwilhelm