https://github.com/benhid/logmap-embeddings
https://github.com/benhid/logmap-embeddings
Last synced: 10 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/benhid/logmap-embeddings
- Owner: benhid
- License: mit
- Created: 2022-06-23T05:39:01.000Z (almost 4 years ago)
- Default Branch: main
- Last Pushed: 2022-06-28T05:39:40.000Z (almost 4 years ago)
- Last Synced: 2025-03-25T06:23:50.037Z (about 1 year ago)
- Language: Python
- Size: 27.2 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# LogMap-ML Embeddings
[](https://openaccess.city.ac.uk/id/eprint/25810/1/ESWC2021_ontology_alignment_LogMap_ML.pdf)
[](LICENSE)
> Work in Progress!
Check out the parent project [LogMap matcher](https://github.com/ernestojimenezruiz/logmap-matcher/).
## Get started
### Setup
```console
$ python3 -m venv .venv
$ source .venv/bin/activate
$ python -m pip install -r requirements.txt
```
### Usage
#### Pre-process: Run the original system
Run LogMap 4.0:
```console
$ mkdir logmap_output_oaei_22_bioml
$ java -jar logmap/logmap-matcher-4.0.jar MATCHER \
file:$(pwd)/use_cases/oaei_22_bioml/fma.body.owl file:$(pwd)/use_cases/oaei_22_bioml/snomed.body.owl $(pwd)/logmap_output_oaei_22_bioml/ true
```
This leads to LogMap initial set of candidate mappings or _anchors_
(
)
and
over-estimation class mappings
(
).
#### Get Embedding Models
You can either download the word2vec embedding by gensim (the one trained with a corpus of Wikipedia articles from 2018-[download](https://drive.google.com/file/d/1rm9uJEKG25PJ79zxbZUWuaUroWeoWbFR/view?usp=sharing)) or use the ontology-tailored [OWL2Vec\*](https://github.com/KRR-Oxford/OWL2Vec-Star) embedding. The ontologies use one common embedding model.
```bash
$ python deepwalk.py use_cases/oaei_22_bioml/merged_with_mappings.owl --walk-number 10 --walk-length 2 --output deepwalk_model/
```
#### Prepare Dataset
Use the provided [standalone](standalone.py) script:
```bash
$ python standalone.py --cache-dir cache_standalone-dist_bioml --config default_bioml.cfg
```
LogMap-ML will extract the class labels for each class in both ontologies and generate high-confidence train mappings
(_seed mappings_ 
)
for training.
It will also create a samples dataset from a set of high recall candidate mappings (LogMap’s over-estimation mappings ) for evaluation.
#### Evaluate
Assuming that gold standards (complete ground truth mappings) are given, Precision and Recall can be directly calculated by:
```bash
$ python evaluate.py --cache-dir cache_standalone-dist_bioml/ --reference use_cases/oaei_22_bioml/reference.txt --distances cache_standalone-dist_bioml/distances.txt --mappings logmap_output_oaei_22_bioml/logmap2_mappings.txt
```
## Publications
* Jiaoyan Chen, Ernesto Jimenez-Ruiz, Ian Horrocks, Denvar Antonyrajah, Ali Hadian, Jaehun Lee. **Augmenting Ontology Alignment by Semantic Embedding and Distant Supervision**. European Semantic Web Conference, ESWC 2021. ([PDF](https://openaccess.city.ac.uk/id/eprint/25810/1/ESWC2021_ontology_alignment_LogMap_ML.pdf))
## License
This project is licensed under the terms of the MIT - see the [LICENSE](LICENSE) file for details.