Language Models as Knowledge Embeddings

Authors: Xintao Wang, Qianyu He, Jiaqing Liang, Yanghua Xiao

IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that LMKE achieves state-of-the-art performance on KE benchmarks of link prediction and triple classification, especially for long-tail entities. We conduct extensive experiments on widely-used KE benchmarks
Researcher Affiliation Academia 1Shanghai Key Laboratory of Data Science, School of Computer Science, Fudan University 2Fudan-Aishu Cognitive Intelligence Joint Research Center
Pseudocode No No pseudocode or algorithm blocks were found in the paper.
Open Source Code Yes Our codes are available at https://github.com/Neph0s/LMKE
Open Datasets Yes We experiment on four popular benchmark datasets: FB13 [Socher et al., 2013], FB15k237 [Toutanova, 2015], UMLS [Dettmers et al., 2018] and WN18RR [Dettmers et al., 2018], whose statistics are shown in Table 3.
Dataset Splits Yes We experiment on four popular benchmark datasets: FB13 [Socher et al., 2013], FB15k237 [Toutanova, 2015], UMLS [Dettmers et al., 2018] and WN18RR [Dettmers et al., 2018], whose statistics are shown in Table 3. Table 3: Statistics of the datasets. Dataset #Entity #Relation #Train #Dev #Test Avg DL. We search these hyperparameters with BERT-tiny: learning rate of PLMs among {10 4, 5 10 5, 10 5}, learning rate of other components among {10 3, 5 10 4, 10 4, 10 5}, batch size among {12, 16, 32, 64} based on best Hits@10 on the dev set.
Hardware Specification No No specific hardware details (e.g., GPU model, CPU type, memory) were found. The paper mentions models are "fine-tuned" but does not specify the computational resources used.
Software Dependencies No The paper mentions using "BERT-base [Devlin et al., 2018] and BERT-tiny [Turc et al., 2019] as the language model" and "Adam as the optimizer." However, specific version numbers for these or other software dependencies (e.g., Python, PyTorch, TensorFlow, CUDA) are not provided.
Experiment Setup Yes We search these hyperparameters with BERT-tiny: learning rate of PLMs among {10 4, 5 10 5, 10 5}, learning rate of other components among {10 3, 5 10 4, 10 4, 10 5}, batch size among {12, 16, 32, 64} based on best Hits@10 on the dev set. With BERT-base, we set the batch size as 16 for triple classification and 12 for link prediction, which considers both results of BERT-tiny and limited memory. For triple classification, we sample 1 negative triple for each positive triple.