Versions and changes

Upcoming

  • added new parameter reset_annids AnnotatationSet.clear
  • fixed a bug in StringGazetteer when the gazetteer list contains a string only entry instead of a tuple.
  • Added parameters row1_style and row2_style to Document.show
  • Fixed issue 149: reset the annotation id after clearing the annotation set
  • ELG annotator: allow to use a pre-initialized authentification file
  • ELG annotator: add utility function to convert UDPipe annotations: udptoken2tokens
  • Added IBM NLU annotator
  • Added Google NLP annotator
  • Loading a GATE XML using Document.load(path, fmt="gatexml") now supports feature values of type Map, List, Set, Array, Date

1.0.6 (2021-11-19)

  • The minimum Python version has been changed from 3.6 to 3.7. This now allows the use of postponed evaluation of type annotations and the use of dataclasses.
  • add Document.edit(edits, affected_strategy="keepadapt") method: update document text and change offsets/indices for all annotations, if necessary.
  • New annotator StringRegexpAnnotator which allows to annotate documents using Python regular expressions in a very simple and flexible way. See the StringRegexpAnnotator Documentation.
  • The StringGazetteer has been implemented. See the Gazetteers documentation.
  • The TokenGazetteer parameter names got changed to match the corresponding StringGazetteer names
  • ! the parameter name out_set in gatenlp.processing.tokenizer was changed to outset_name to be consistent with the name used elsewhere.
  • ! the parameter name out_annset in gatenlp.processing.client was changed to outset_name to be consistent with the name used elsewhere.
  • The Document.clone() method can be used to easily create an exact copy of a document, where none of the data is shared (deep copy)
  • The TextNormalizer has been added. It can be used to normalize the unicode representation of the text in a document.
  • loading a document from bdocjs format now does not require any keys in the JSON map and also ignores all unknown keys. This allows to more easily import ad-hoc documents which e.g. only contain the text or text and annotations (if not offset type is specified, python is assumed).
  • The documentation has been updated and extended (especially for gazetteers and PAMPAC)

1.0.5.1 (2021-10-09)

  • Bug fix: make lib_spacy support both versions 2.x and 3.x (1.0.5 used a method which is only available in 3.x)

1.0.5 (2021-10-08)

Changes that break backwards compatibility:

  • AnnotationSet.with_type() previously returned a detached set with all annotations if no types were specified, this now returns a detached set with no annotations which is more logical.
  • API changes:
  • pam.pampac.actions.AddAnn: parameter anntype has been changed to type
  • The Feature() constructor kw arg logger has been changed to _change_logger and deepcopy has been changed to _deepcopy
  • Pampac: use the term "matches" instead of "data" for the named information stored for each named pattern that fits the document. A single one of these is often called "match info" and the index for a specific info is now called "matchidx" instead of "dataidx". See issue #89
  • Parameter spacetoken_type for AnnSpacy and spacy2gatenlp has been changed to space_token_type to conform to the parameter name used for AnnStanza and stanza2gatenlp.
  • Stanford Stanza support now requires Stanza version 1.3.0 or higher
  • Changes to lib_spacy: new parameter containing_anns to apply the spacy pipeline only to the part of the document covered by each of the annotations in the annotation set or iterator. New parameters component_cfg to specify a component config for Spacy and retrieve_spans to retrieve additional span types to retrieve.
  • Several bugfixes in Pampac.

Other changes and improvements:

  • New method AnnotationSet.create_from(anniterable) to create a detached, immutable annotation set from an iterable of annotations
  • New method Document.anns(annspec) creates a detached set of all annotations that match the specification
  • New method Document.yield_anns(annspec) yields all annotations which match the specification
  • Fixed bug in Token Gazetteer: issue #93
  • Pampac: there is now a PampacAnnotator class to simplify using Pampac in a pipeline.
  • Pampac: New parameter containing_anns for Pampac.run: if specified, runs the rules on each span of each of the containing annotations
  • Pampac: a Result is now an Iterable of match infos.
  • Pampac: the .within(..) .contains(..) etc. constraints now allow to use a separate annotation set, e.g. .within("Person", annset=doc.annset("Other")). See issue #57
  • Pampac: RemoveAnn action has been added
  • Pampac: UpdateAnnFeatures has been improved
  • Pampac: AddAnn action supports getter helpers in feature values
  • Span objects are now immutable. Equality and hashing of Span objects are based on their start and end offsets.
  • Annotation equality and hashing has been changed back to the Python default: variables compare only equal if they reference the same object and hashing is based on object identity. For comparing annotations by content, the methods ann.equal(other) (compare content without annotation id) and ann.same(other) (compare content including annotation id) have been implemented.
  • Documents can be saved in "tweet-v1" format
  • Fixed a problem with the HTML viewer: leading and multiple whitespace annotations now show correctly.

1.0.4 (2021-04-10)

  • The GateWorkerAnnotator parameters have been changed: instead of parameters gatehom and port, the parameter gateworker now needs to receive a GateWorker instance. Also the update_document parameter has been added and now allows both updating and replacing the Python document from the Java GATE document
  • Issue #66: make it possible to show annotations over new-lines in the html ann viewer
  • Issue #65: provide ParagraphTokenizer and SplitPatternTokenizer to easily annotate paragraphs and other spans separated by some split pattern
  • Issue #73: pickle document with offset index created
  • Issue #68: rename the main development branch from "master" to "main"
  • Issue #74: fix a bug in PAMPAC related to matching an annotation after some text
  • Various improvements, additions and bug fixes in Pampac
  • Issue #75: GateWorker now shows any Java exception when starting the Java process fails
  • Issue #76: GateWorker has a new method loadPipelineFromUri(uri)
  • Issue #77: GateWorkerAnnotator now automatically loads a pipeline from a URL if the string passed to the pipeline parameter looks like a URL or if it is the result of urllib.parse.urlparse. It is always treated like a file if it is a pathlib.Path
  • added the Actions action for Pampac to recursively wrap several actions into one
  • allow each Rule to have any number of actions, change signature to Rule(patter, *actions, priority=0)
  • The Pampac AddAnn action does not require a value for the name parameter any more, if not specified, the full span of the match is used.
  • New method add_anns(anniterable) to add an iterable of annotations to a set
  • The document viewer now also works in Google Colab
  • The GateWorker can now be used as context manager: with GateWorker() as gw:

1.0.3.1 (2021-03-01)

  • add training course slides
  • fix issue #63: could not import html document from a local file

1.0.3 (2021-02-22)

  • Fix issues with logging and error handling in executor module
  • Improve/add/change document sources/destination JsonLinesFile
  • add Span.embed method
  • Implement multi-word tokens (MWTs) for the Stanza annotator
  • Add support for space tokens for the Stanza annotator
  • Support showing annotations over trailing spaces in the html ann viewer
  • Add the Document.attach(annset) method (mostly for internal use only!)
  • Add the ConllUFileSource to import CoNLL-U corpora
  • Fix a problem in the html ann viewer where unnecessary spans were created
  • Add option to the Document.show() method to style the document text div

1.0.2 (2021-02-09)

  • Fix issue #56: Rename GateSlave to GateWorker

1.0.1 (2021-02-07)

  • Initial release