Skip to content

Releases: kuhumcst/DanNet

2024-08-09: Supersenses + OEWN update

09 Aug 09:53
Compare
Choose a tag to compare
  • Added 71055 Supersenses to DanNet, mostly based on a mapping devised for the SemDaX corpora.
    • Around 300 of these were "fixed" according to hypernym, differing from the mapping.
    • Another 3793 of these were created based on a new mapping of unmapped synsets.
  • Added sentiment to 183 synsets taken from the old DanNet 2 connotation data.
  • The existing links to the 2022 edition of the Open English WordNet were replaced with links to the 2023 edition (new synset IDs).
  • The OEWN extension dataset has been updated such that it uses the new OEWN synset IDs.

ELEXIS-WSD 1.1 with supersenses (PARTIAL)

08 Aug 10:58
Compare
Choose a tag to compare

NOTE: THIS IS NOT A DANNET RELEASE!

The partially supersense-annotated Danish ELEXIS-WSD 1.1 corpus.

All tokens with DanNet sense IDs have been annotated, but the remaining tokens have not. See supersenses-discussion.md for more.

See issues: #138, #141, #144, #145

2024-06-12: WN-LMF as an alternative format

12 Jun 10:57
Compare
Choose a tag to compare
  • WN-LMF has been added as an alternative format following a request on Github. The new file, dannet-wn-lmf.xml.gz, can even be used directly in software such as goodmami/wn (see also: example on Github). Unfortunately, WN-LMF currently does not support the full set of data found in DanNet; for instance, our ontological types are not present in this format and the same applies to DanNet-specific relations such as used for.
  • A total of 1906 bad source references to DDO have been removed from the dataset. These dns:source-relations had been created automatically based on IDs that exclusively exist within DanNet and for this reason they couldn't reference DDO.
  • 88 Synset definitions have been fixed such that the split between titles and occupations is correct.
  • Synset indegrees have been recalculated.

2024-04-30: Improved CSV export + other small fixes

30 Apr 07:39
Compare
Choose a tag to compare
  • The CSV export has been improved by...
    1. ... removing the presence of internal IDs in synsets.csv (referring to ontological types) replacing them instead with the concrete mix of ontological types.
    2. ... including lexical entries in words.csv which were previously erroneously excluded.
  • Some Lexical entries which were formerly of the generic ontolex:LexicalEntry type now have more specific types, e.g. ontolex:Word, ontolex:MultiWordExpression, or ontolex:Affix.
  • Some of the parts-of-speech for the adjectives added in release 2023-05-11 were missing a PoS relation and/or mixed up two separate relation types; this has now been fixed.

2023-11-28: Short labels

28 Nov 10:42
Compare
Choose a tag to compare

CSV NOT AVAILABLE - THIS RELEASE IS MISSING A CSV EXPORT!

  • dns:shortLabel variants of synset labels (derived from, amongst other things, word frequencies from DDO) have been added to the DanNet dataset.
  • dns:source is now used once again to link to the original dictionary entry sources such as DDO. The usage of dc:source was both problematic wrt. its definition in the schema, as well the annoying fact that dc in some cases results in confusion when used as an RDF prefix as it may be hardcoded to a specific IRI.
  • Some sense labels had lost their language (da) by mistake and this has now been fixed.

NOTE: I've also included synset-indegrees.edn which is used internally by the DanNet web app for sorting and visualisations (e.g. word clouds). Furthermore, the data is no longer duplicated into a union dataset, since the end user can simply load multiple datasets into the desired graph.

2023-09-28: Fixing the domain topic relation

28 Sep 11:47
Compare
Choose a tag to compare
  • The wn:has_domain_topic relation had been used in place of wn:domain_topic in the DanNet dataset. This has now been corrected.

2023-07-07: Thousands of new links and schema updates

07 Jul 09:35
Compare
Choose a tag to compare
  • DanNet now has around 10K new links to the CILI which is also linked with the OEWN and other WordNets.
  • Some new relations (dns:eqHypernym, dns:eqHyponym, and dns:eqSimilar) have been added since wn:ili and wn:eqSynonym were not sufficient to cover the inter-WordNet links we now have.
  • DanNet synsets now also have the raw DDO domain values from DSL that were present in the older versions of DanNet. These are represented with the dc:subject relation.
  • Furthermore, the sex/gender data from the older versions of DanNet has also been included. It is available via the new dns:gender relation.
  • To better facilitate graph navigation on the DanNet website, a new relation called dns:linkedConcept has been added to the DanNet schema. This relation is the inverse of wn:ili and is inferred in the large graph that can be queried on wordnet.dk/dannet.

2023-06-01: Partial inclusion of English links

19 Jun 12:07
Compare
Choose a tag to compare
  • The schema translations have been updated.
  • Around 5000 links have been added which link DanNet to the Open English WordNet or indirectly via the CILI.
  • The OEWN data set has received a companion data set containing generated labels for synsets, senses, and words.
  • dns:dslSense and dns:source have been removed from the DanNet schema (dns:source has been replaced by dc:source)

The actual release date was put back till now nice since my wife went into labour on the planned release date.

2023-05-23: DDS/COR improvements & links to DDO

23 May 11:09
Compare
Choose a tag to compare

Changelog

  • Many DanNet words and senses have been linked to DDO via the new dns:source relation.
  • Unofficial conjugations present in the COR companion data set have been marked as such in their rdfs:label.
  • Various other smaller tweaks to the COR dataset.
  • The DDS data set now uses 32-bit float as opposed to double, which results in a smaller RDF export as this data type doesn't require any special encoding in .ttl-files.

2023-05-11: The new DanNet

23 May 09:13
Compare
Choose a tag to compare

This is the first proper release of the new RDF-native DanNet! 🥳

Please note that any hyperlinks in the description below will resolve to whatever the current version of DanNet is, not this particular release which is instead available in the attached .zip-files. This Github repository currently serves as an archive of the DanNet releases.

Overview of data

DanNet is based on the Ontolex standard with additions from the Global WordNet Association. You can explore DanNet directly on wordnet.dk, but you may also download our data as an RDF data set or in a slightly more limited CSV edition. All our data sets are published under the CC BY-SA 4.0 license.

DanNet is integrated with COR and DDS too. These alternative RDF data sets can also be downloaded on this page (COR integration og DDS integration).

You may also download DanNet+COR+DDS as a combined data set or download a complete copy of all the data that can be found on wordnet.dk/dannet, including logically inferred data and associated RDF schemas.

Changelog

There are too many changes in this initial release to list all of them in a succinct way:

  • Around 5000 new senses have been added, mostly adjectives.
  • Many data set inconsistencies and other undesirable properties have been cleaned up.
  • The entirety of DanNet has been converted to the Ontolex standard and uses the relations from the Global WordDet Association.
  • DanNet is now RDF-native; RDF schemas are also available covering e.g. the ontological types.
  • The DSL-derived DanNet IDs all resolve to actual RDF resources which can be viewed in a browser.
  • Several companion data sets are available for download and are also merged with the data on wordnet.dk.
  • Additional data points have also been inferred from the bootstrap data, e.g. inverse relations.
  • The CSV download is now CSVW and includes metadata files describing contents of the columns.
  • The DanNet data is now licensed as CC BY-SA 4.0 and the source code of the project is available under the MIT licence.
  • ... and of course wordnet.dk/dannet is the new home of DanNet.

In addition to the work done on DanNet itself, we have also contributed to the Global WordNet Association's RDF schema.