Keci embedding models for explainable benchmarking (QALD10, QALD9+DB, QALD9+WK)

This project provides embedding models for Drill. The embeddings are generated for three knowledge graphs created for the three QA benchmarks QALD 9 plus DBpedia, QALD 9 plus Wikidata and QALD 10. Here are some more details: Three knowledge graphs built for QALD 9 plus DBpedia, QALD 9 plus Wikidata and QALD10 can be found here. The following table gives an overview of the parameters, that we use to calculate the Keci embeddings. Parameter QALD9+DB QALD9+WD QALD10 Model Keci Keci Keci Dimensions 8 8 32 Learning rate 0.1 0.1 0.1 Epochs 1 1 1 Batch size 512 512 50000 Optimization function Adam Adam Adam Evaluation mode None None None Complete command dicee --dataset_dir KGs/QALD9_DB --model Keci --embedding_dim 8 --lr 0.1 --save_embeddings_as_csv --num_epochs 1 --batch_size 512 --optim Adam --scoring_technique NegSample --eval_mode None dicee --dataset_dir KGs/QALD9_WK --model Keci --embedding_dim 8 --lr 0.1 --save_embeddings_as_csv --num_epochs 1 --batch_size 512 --optim Adam --scoring_technique NegSample --eval_mode None dicee --dataset_dir KGs/QALD10 --model Keci --embedding_dim 32 --lr 0.1 --save_embeddings_as_csv --num_epochs 1 --batch_size 50000 --optim Adam --scoring_technique NegSample --eval_mode None