Skip to content

Latest commit

 

History

History
55 lines (40 loc) · 1.35 KB

README.md

File metadata and controls

55 lines (40 loc) · 1.35 KB
Raw

DBLP Entity Linker

Demo is hosted on https://ltdemos.informatik.uni-hamburg.de/dblplink.

Download pre-trained models from http://ltdata1.informatik.uni-hamburg.de/debayan-dblplink/ and untar them in api/.

Docker

docker compose build
docker compose up

At the moment, the Elasticsearch dump for the KG embeddings exceeds 100 GB in size, hence we do not upload it in public. Kindly contact us if you wish to receive the dumps individually.

Training label span detector:

cd extras/
python -u train.py --model_name t5-small --batch_size 4 --epochs 10 --lr 0.001

Evaluating label span detector:

cd extras/
python -u evaluate.py --model_name t5-base --embedding_name transe

Training re-ranker model

cd reranker/
python siamese.py data_distmult/train.jsonlines data_distmult/valid.jsonlines output_model_dir/

Running Streamlit UI

streamlit run Home.py staging

OR

streamlit run Home.py production

Citation

@inproceedings{10.1145/3448016.3457280,
	author = {Banerjee, Debayan and , Arefa and Usbeck, Ricardo and Biemann, Chris},
	title = {DBLPLink: An Entity Linker for the DBLP Scholarly Knowledge Graph},
	year = {2023},
	booktitle = {The Semantic Web – ISWC 2023: 22nd International Semantic Web Conference, Athens, Greece, November 6–10, 2023, Proceedings},
	location = {Athens, Greece}
}