Releases: korpling/annatto
Releases · korpling/annatto
0.9.0 - 2024-06-24
Release Notes
Added
link, map, enumerate, and collapse have documentation visible to the user.
Fixed
- documentation for import of
xlsx showed wrong config doc string
link does not use default 0 for source_node and target_node attributes anymore, since they are 1-based indices (instead, there is no default)
Download annatto 0.9.0
0.8.2 - 2024-06-21
Release Notes
Fixed
sequence export for horizontal data now also works in models with multiple segmentation and empty tokenscheck can now save without a panic when report attribute is omitted. list is the default report level which only applies to save, not to the report attribute itself, where the default is not to print.
Download annatto 0.8.2
0.8.1 - 2024-06-21
Release Notes
Fixed
sequence export for horizontal mode now works
Download annatto 0.8.1
0.8.0 - 2024-06-17
Release Notes
Added
- Importer for the relANNIS format (http://korpling.github.io/ANNIS/3.7/developer-guide/annisimportformat.html)
- progress reports for
enumerate, link, and map
revise can now rename nodes using attribute node_names, e. g. for renaming (top level) corpus nodes. The syntax is equivalent to renaming annotations, thus renaming with an empty value will lead to deletion. Renaming with an existing value (also rename with self) will lead to an error.- Add
zip option to GraphML export to directly export as ZIP file which can be
more easily imported in ANNIS.
Changed
- update to dependencies to latest graphANNIS version
Fixed
- Fix non-resolved relative path when importing EXMARaLDA files.
- Limit the table width when listing the module properties, so they fit in the
current terminal.
Download annatto 0.8.0
0.7.0 - 2024-05-23
Release Notes
Added
sequence exports connected node's annotation values (e. g. ordered nodes) as vertical or horizontal sequences.split breaks up conflated annotation values into partsrevise now offers to delete an entire subgraph from a node in the inverse direction of part of edgesenumerate can prefix the numeric annotation it generates with an annotation value from the query match (use attribute value to point in the match list with a 1-based index)
Changed
enumerate uses u64 internally (to be in line with graphANNIS and to be deserializable)collapse now uses node ids that indicate the node names that entered the merge, the parent node is not indicated anymoresplit has default configuration/behaviour (do nothing); attribute keep is now delete to adhere to boolean default logic
Fixed
- no more
annis::tok labels for non-terminal coverage nodes in xlsx import
- hypernode id's are unified, in older versions it could happen that annotations get distributed about two or more hypernode instances due to invalid determination of the parent (part of-child)
Download annatto 0.7.0
0.6.0 - 2024-04-22
Release Notes
Added
- Added simple chunker module based on
text-splitter.
check can write check report to filecheck can test a corpus graph comparing results to an external corpus graph loaded from a graphANNIS database- import
ptb can now split node annotations to derive a label for the incoming edge, when a delimiter is provided
using edge_delimiter. E. g., NP-sbj will create a node of category NP, whose incoming edge has function sbj,
given the following config is used: edge_delimiter = "-"
- config attribute
stable_order for exporting graphml enforces stable ordering of edges and nodes in output
- toml workflow files now strictly need to stick to known fields of module structs
- command line interface now has the
list subcommand to list all modules and the info subcommand to show the description and parameters of a module.o
Changed
- The
check module can now query the AnnotationGraph directly without using
the CorpusStorageManager.
chunk deserializes with empty config to default values
Fixed
- Don't throw error if output directory for any workflow does not exist.
- import
ptb: Also constituents get PartOf edges to their respective document node.
Download annatto 0.6.0
0.5.0 - 2024-01-19
Release Notes
Changed
- improve progress reporting by reporting each conversion step separately
Changed
- improve progress reporting by reporting each conversion step separately
Added
- graph_op
collapse can collapse an edge component, i. e., it merges all nodes in a connected subgraph in said component
collapse can be accelerated when all edges of the component to be collapsed are known to be disjoint by providing disjoint = true in the step configcollapse provides more feedback on current processcollapse gives hypernodes proper names that allow to identify the subgraph they belong to. Furthermore already existing hypernode ids are not reused (in case multiple collapse operations are run on a graph).CorpusStorage is now quiet- importing
exmaralda does now has more features
exmaralda can be exportedxlsx import creates part of-edges between tokens and document nodes- all imports add PartOf edges from nodes to their respective document (lowest corpus node)
Fixed
link now considers all matching nodes for the same value, so the correct amount of edges is createdexmaralda returns error when there is no time value for a timeline item- fixed and simplified import of corpus node annotations
exmaralda import's paths to linked media files are relative to the working directoryxlsx importer now adds PartOf relations to the document nodes
Download annatto 0.5.0
0.4.0 - 2023-11-13
Release Notes
Added
- a separator for joining node values in
link can be set with attribute value_sep
- spreadsheet imports can now be configured with a fallback token column for annotation names not mentioned in a column map, an empty string means map to timeline directly
- graph_op
check can now be configured to not let the entire processing chain fail, when a test fails, by setting policy = "warn" (default is fail)
- metadata can be imported from spreadsheets alongside the linguistic data in the workbook, a data and a metadata spreadsheet name or number can now be specified for importing xlsx
- add heuristic for KWIC visualizer in graphml export
re is now reviserevise can modify componentspath as a import format now triggers the embedding of path names as nodes into the graph; this is supposed to help to represent configuration files for ANNIS- import
path adds an annis::file annotation
- import
path adds part-of edges
- very basic implementation of a generic xml importer
- import opus sentence alignments
- graph op
enumerate to enumerate nodes, i. e., add numeric annotations to results of one or multiple queries
- add importer for the format used by the TreeTagger
Fixed
- mapping annotations now correctly extracts the id of the node to apply a new annotation to
- linking nodes failed to extract node names when graphANNIS responded with a node name only (e. g. in case of "tok" or "node" in a query)
- linking nodes did not concatenate the values of multiple nodes properly, this is now fixed
- fixed code of spreadsheet import (merged cells might not have an end column reference)
- relative import and export paths are interpreted as relative to the parent directory of the workflow file
- the spreadsheet importer will use the correct namespace
default_ns for segmentation ordering relations
- fixed ordering of token nodes in spreadsheet import
Removed
- removed
show-documentation subcommand and moved the documentation from mdBook to the crate documentation in the source code
Download annatto 0.4.0
[0.3.1] - 2023-08-04
Release Notes
Fixed
- Documentation was not included in release binaries.
Download annatto 0.3.1
[0.3.0] - 2023-08-04
Release Notes
Changed
- CLI binary renamed from
annatto-cli to annatto
- To execute a workflow file, use
annatto run <workflow-file>
- module properties are now struct attributes of the importer, manipulator, or exporter, which facilitates deserialization and also use (undefined properties are no longer accepted, required properties cannot be ommited)
- only TOML workflow files are now supported, xml workflows can no longer be processed
- TOML support adds a command
annatto validate <workflow-file> that checks if a worklow description can be deserialized to an internal workflow
- not all modules have a default implementation anymore (path attributes have no default value that makes sense)
- there is a default operation for each step type. Import: Create empty corpus, Manipulation: Do nothing, Export: Write GraphML
- map properties of some modules (such as
tier_group for importing Textgrid) are no longer String codings, since TOML supports providing maps directly
- flattened TOML for workflow files
- TOML workflows: module config has to be singled out in separate table
check tests are now configured in main workflow as TOML fragmentcheck report table contains number of matches in case of failure- linker takes list of node indices for value nodes (source and target)
Added
- collected errors in status messages
Failed are now all reported at the end
of the job
- an annotation mapper can create annotations from existing annotations using
AQL for defining target nodes
- New command
show-documentation for CLI, which starts a browser with the user
guide.
- after running
check, the the test results can be printed as a table (default: off)
check displays matching nodes for tests in new verbose modecheck now comes with a higher level test ("Layer test") that is internally converted into atomic aql tests. The test can be applied to nodes and edges. It tests if a layer exists and only valid annotation values have been used.- using flag
--env allows to resolve environmental variables in workflow definitions which enables the use of template workflow definitions
- node linker: with two queries the resulting nodes can be linked via edges of a configurable type, layer, and name
- boolean environment variable
ANNATTO_IN_MEMORY influences whether or not graphs will be stored on disk or in memory
Fixed
- fixed panics caused by undefined attributes in tier tag or missing speaker table / wrong speaker id
- exmaralda import did not properly forward errors through the status sender, which it now does
Download annatto 0.3.0
