Hierarchy Transformers (HiTs)

Code repository for the paper: "Language Models as Hierarchy Encoders".

Installation

Main Dependencies

This repository follows a similar layout as the Sentence Transformers library. It mainly depends on the following libraries:

Sentence Transformers for language models.
DeepOnto for processing hierarchies and constructing datasets from hierarchies.
Geoopt for arithmetic in hyperbolic space.

Install from PyPI

# requiring Python>=3.8
pip install hierarchy_transformers

Install from GitHub

pip install git+https://github.com/KRR-Oxford/HierarchyTransformers.git

Models on Huggingface Hub

Our HiT models are released on the Huggingface Hub.

Get Started

Use the following code to get started with HiTs:

from hierarchy_transformers import HierarchyTransformer
from hierarchy_transformers.utils import get_torch_device

# set up the device (use cpu if no gpu found)
gpu_id = 0
device = get_torch_device(gpu_id)

# load the model
model = HierarchyTransformer.load_pretrained('Hierarchy-Transformers/HiT-MiniLM-L12-WordNet', device)

# entity names to be encoded.
entity_names = ["computer", "personal computer", "fruit", "berry"]

# get the entity embeddings
entity_embeddings = model.encode(entity_names)

Default Probing for Subsumption Prediction

Use the entity embeddings to predict the subsumption relationships between them.

# suppose we want to compare "personal computer" and "computer", "berry" and "fruit"
child_entity_embeddings = model.encode(["personal computer", "berry"], convert_to_tensor=True)
parent_entity_embeddings = model.encode(["computer", "fruit"], convert_to_tensor=True)

# compute the hyperbolic distances and norms of entity embeddings
dists = model.manifold.dist(child_entity_embeddings, parent_entity_embeddings)
child_norms = model.manifold.dist0(child_entity_embeddings)
parent_norms = model.manifold.dist0(parent_entity_embeddings)

# use the empirical function for subsumption prediction proposed in the paper
# `centri_score_weight` and the overall threshold are determined on the validation set
# see source code at `src/hierarchy_transformers/evaluation` for more details about our implementation for the hyperparameter tuning.
subsumption_scores = - (dists + centri_score_weight * (parent_norms - child_norms))

Datasets

Datasets for training and evaluating HiTs are available at Zenodo, including those constructed from:

WordNet
SNOMED CT
Schema.org
FoodOn
DOID

License

!!! license "License"

Copyright 2023 Yuan He.
All rights reserved.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at *<http://www.apache.org/licenses/LICENSE-2.0>*

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

Citation

The preprint of our paper for is currently available at arxiv.

Yuan He, Zhangdie Yuan, Jiaoyan Chen, Ian Horrocks. Language Models as Hierarchy Encoders. arXiv preprint arXiv:2401.11374 (2024).

@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}
}

KRR-Oxford / HierarchyTransformers