Datasets:
child
stringlengths 1
71
| parent
stringlengths 1
71
| label
int64 0
1
|
---|---|---|
boarhound | hound | 1 |
boarhound | beagle | 0 |
boarhound | afghan hound | 0 |
boarhound | staghound | 0 |
boarhound | wolfhound | 0 |
boarhound | ibizan hound | 0 |
boarhound | plott hound | 0 |
boarhound | foxhound | 0 |
boarhound | norwegian elkhound | 0 |
boarhound | weimaraner | 0 |
boarhound | redbone | 0 |
chin | feature | 1 |
chin | jaw | 0 |
chin | jowl | 0 |
chin | cheek | 0 |
chin | brow | 0 |
chin | temple | 0 |
chin | hydremia | 0 |
chin | deserter | 0 |
chin | suppressor | 0 |
chin | tilefish | 0 |
chin | damp | 0 |
fertilization | creation | 1 |
fertilization | recess | 0 |
fertilization | walk-through | 0 |
fertilization | ileocolic vein | 0 |
fertilization | kopek | 0 |
fertilization | poon | 0 |
fertilization | elasticity | 0 |
fertilization | summer school | 0 |
fertilization | water vapor | 0 |
fertilization | billiard ball | 0 |
fertilization | fugue | 0 |
hulk | ship | 1 |
hulk | treasure ship | 0 |
hulk | whaler | 0 |
hulk | abandoned ship | 0 |
hulk | gas-turbine ship | 0 |
hulk | school ship | 0 |
hulk | hospital ship | 0 |
hulk | icebreaker | 0 |
hulk | pirate | 0 |
hulk | troopship | 0 |
hulk | passenger ship | 0 |
change of location | movement | 1 |
change of location | periodic motion | 0 |
change of location | crustal movement | 0 |
change of location | turning | 0 |
change of location | recoil | 0 |
change of location | brownian movement | 0 |
change of location | throw | 0 |
change of location | wave | 0 |
change of location | passing | 0 |
change of location | twist | 0 |
change of location | bending | 0 |
pheasant | wildfowl | 1 |
pheasant | quail | 0 |
pheasant | grouse | 0 |
pheasant | partridge | 0 |
pheasant | radar echo | 0 |
pheasant | transfusion | 0 |
pheasant | deep-freeze | 0 |
pheasant | scotch kiss | 0 |
pheasant | tollbooth | 0 |
pheasant | bevatron | 0 |
pheasant | genus tragopan | 0 |
posterior synechia | synechia | 1 |
posterior synechia | anterior synechia | 0 |
posterior synechia | humanitarianism | 0 |
posterior synechia | acerola | 0 |
posterior synechia | sericea lespedeza | 0 |
posterior synechia | religious orientation | 0 |
posterior synechia | point | 0 |
posterior synechia | pulp | 0 |
posterior synechia | airfoil | 0 |
posterior synechia | altruism | 0 |
posterior synechia | alliaria | 0 |
paradoxical sleep | sleep | 1 |
paradoxical sleep | sleeping | 0 |
paradoxical sleep | shuteye | 0 |
paradoxical sleep | orthodox sleep | 0 |
paradoxical sleep | acanthuridae | 0 |
paradoxical sleep | shrike | 0 |
paradoxical sleep | urticaceae | 0 |
paradoxical sleep | fabrication | 0 |
paradoxical sleep | transition | 0 |
paradoxical sleep | pretension | 0 |
paradoxical sleep | stretcher | 0 |
referred pain | pain | 1 |
referred pain | sting | 0 |
referred pain | causalgia | 0 |
referred pain | glossalgia | 0 |
referred pain | pang | 0 |
referred pain | pleurodynia | 0 |
referred pain | stitch | 0 |
referred pain | distress | 0 |
referred pain | throb | 0 |
referred pain | tenderness | 0 |
referred pain | hemorrhoid | 0 |
diagnostician | specialist | 1 |
Dataset Card for WordNetNoun
This dataset is a collection of Multi-hop Inference and Mixed-hop Prediction datasets created from WordNet's subsumption (hypernym) hierarchy of noun entities for training and evaluating hierarchy embedding models.
- Multi-hop Inference: This task aims to evaluate the model’s ability in deducing indirect, multi-hop subsumptions from direct, one-hop subsumptions, so as to simulate transitive inference.
- Mixed-hop Prediction: This task aims to evaluate the model’s capability in determining the existence of subsumption relationships between arbitrary entity pairs, where the entities are not necessarily seen during training. The transfer setting of this task involves training models on asserted training edges of one hierarchy testing on arbitrary entity pairs of another.
See our published paper for more detail.
Links
- GitHub Repository: https://github.com/KRR-Oxford/HierarchyTransformers
- Huggingface Page: https://huggingface.co/Hierarchy-Transformers
- Zenodo Release: https://doi.org/10.5281/zenodo.10511042
- Paper: Language Models as Hierarchy Encoders (NeurIPS 2024).
The information of original entity IDs is not available in the Huggingface release; To map entities back to their original hierarchies, refer to this Zenodo release.
Dataset Structure
Each subset in this dataset follows the naming convention TaskType-NegativeType-SampleStructure
:
TaskType
: EitherMultiHop
orMixedHop
, indicating the type of hierarchy evaluation task.NegativeType
: EitherRandomNegatives
orHardNegatives
, specifying the strategy used for negative sampling.SampleStructure
: EitherTriplets
orPairs
, indicating the format of the samples.- In
Triplets
, each sample is structured as(child, parent, negative)
. - In
Pairs
, each sample is a labelled pair(child, parent, label)
, wherelabel=1
denotes a positive subsumption andlabel=0
denotes a negative subsumption.
- In
For example, to load a subset for the Mixed-hop Prediction task with random negatives and samples presented as triplets, we can use the following command:
from datasets import load_dataset
dataset = load_dataset("Hierarchy-Transformers/WordNetNoun", "MixedHop-RandomNegatives-Triplets")
Dataset Usage
For evaluation, the
Pairs
sample structure should be adopted, as it allows for the computation of Precision, Recall, and F1 scores.For training, the choice between
Pairs
,Triplets
, or more complex sample structures depends on the model's design and specific requirements.
Citation
The relevant paper has been accepted at NeurIPS 2024 (to appear).
@article{he2024language,
title={Language models as hierarchy encoders},
author={He, Yuan and Yuan, Zhangdie and Chen, Jiaoyan and Horrocks, Ian},
journal={arXiv preprint arXiv:2401.11374},
year={2024}
}
Contact
Yuan He (yuan.he(at)cs.ox.ac.uk
)
- Downloads last month
- 115