Relation Extraction Using Supervision from Topic Knowledge of Relation Labels

Authors: Haiyun Jiang, Li Cui, Zhe Xu, Deqing Yang, Jindong Chen, Chenguang Li, Jingping Liu, Jiaqing Liang, Chao Wang, Yanghua Xiao, Wei Wang

IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments to evaluate the proposed framework and observe improvements in AUC of 11.5% and max F1 of 5.4% over the baselines with state-of-the-art performance.
Researcher Affiliation Academia 1Shanghai Key Laboratory of Data Science, School of Computer Science, Fudan University, China 2School of Data Science, Fudan University, Shanghai, China 3Shanghai Institute of Intelligent Electronics & Systems, Shanghai, China
Pseudocode No No figure, block, or section explicitly labeled "Pseudocode", "Algorithm", or "Algorithm X" was found. The method is described in narrative text and diagrams.
Open Source Code No The paper does not include an unambiguous statement that the authors are releasing their source code, nor does it provide a direct link to a code repository for the described methodology.
Open Datasets Yes NYT [Riedel et al., 2010] is a distantly supervised RE dataset and is generated by aligning Freebase relations with the New York Times corpus.
Dataset Splits Yes There are 522,611 labeled sentences, 281,270 entity pairs, and 18,252 relational facts (i.e., head entity, relation, tail entity ) in the training set; and 172,448 sentences, 96,678 entity pairs and 1,950 relational facts in the test set.
Hardware Specification No No specific hardware details such as GPU models, CPU models, or memory amounts used for running experiments were provided. The paper only mentions that experiments use embeddings and don't specify the hardware for training.
Software Dependencies No No specific software components with version numbers (e.g., programming language libraries, frameworks, or solvers) are listed as dependencies for reproducibility. It mentions using embeddings and a loss function, but not specific software versions.
Experiment Setup Yes In all experiments, we use the word and position embeddings trained by [Lin et al., 2016] with word dimension d1 = 50 and position dimension d2 = 5. During training, the dropout between layers is used with a dropout rate of 0.1. The loss function is minimized by mini-batch gradient descent, where the batch size is 50, and the learning rate is 0.005. The optimal settings are c = 50, l = 7 and d = 70.