File size: 2,800 Bytes
b81fa1b 405b8b1 b81fa1b b95b987 c35a930 b95b987 b81fa1b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 |
---
language:
- uk
size_categories:
- 100K<n<1M
task_categories:
- question-answering
- summarization
---
# Ukrainian Hypernymy Pairs Dataset
## Background
Hypernymy is the super-subordinate or ISA semantic relation that links more general terms to more specific ones. For example, *rose* is a hyponym of *flower*, a hypernym of *rose*. Words that are hyponyms of the same hypernym are called co-hyponyms, for instance, *rose* and *tulip*. Hyponymy relation is transitive and asymmetric.
Hypernymy is also differentiated by:
* Types — common nouns: *armchair* is a type (hyponym) of *chair*;
* Instances — specific persons, countries, and geographic entities: *Dnipro river* is an instance (instance hyponym) of *river*.
## Project Description
The Ukrainian Hypernymy Pairs Dataset is a collection of noun pairs that express hypernymy relations between words in the Ukrainian language. The dataset contains pairs of words linked by four different types of relations: hypernym-hyponym, co-hyponyms, hypernym-instance, and co-instances.
An example of such a dataset in English may be [BLESS](https://sites.google.com/site/geometricalmodels/shared-evaluation). However, their concepts are linked by one of the following six relations: co-hyponyms, hypernyms, meronyms, attributes, events, and random. Moreover, the hypernymy relation is not divided by terms and instances.
Ukrainian Hypernymy Pairs were constructed utilizing the linkage between [Princeton WordNet](https://wordnet.princeton.edu/), [Wikidata](https://www.wikidata.org/wiki/Wikidata:Main_Page), and Ukrainian [Wikipedia](https://www.wikipedia.org/). We used the Python [Wn package](https://wn.readthedocs.io/en/latest/) to get the relation, which provides an interface to WordNet data.
All terms in the dataset are Wikipedia article titles, and no preprocessing was applied. Therefore it sometimes contains additional info in the brackets.
## Dataset Statistics
This table presents the number of word pairs obtained for each relation type.
| **Relation Type** | **# of Pairs** |
|-----------------------|----------------|
| **Hypernym-Hyponym** | 6,906 |
| **Co-Hyponyms** | 42,860 |
| **Hypernym-Instance** | 2,971 |
| **Co-Instances** | 22,927 |
| **Total # of Pairs** | 275,664 |
## Intended Use
The dataset produced can be particularly valuable for the Hypernym Detection task, where the pair of words is presented to a model, and it should classify whether they are in a hypernymy relation. Other lexico-semantic relations can be added to improve the diversity of the dataset.
## License
Copyright: [Nataliia Romanyshyn](https://twitter.com/supersubnat), [Dmytro Chaplynskyi](https://twitter.com/dchaplinsky), [lang-uk project](https://lang.org.ua), 2023 |