Cross-Relation Cross-Bag Attention for Distantly-Supervised Relation Extraction
Authors: Yujin Yuan, Liyuan Liu, Siliang Tang, Zhongfei Zhang, Yueting Zhuang, Shiliang Pu, Fei Wu, Xiang Ren419-426
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments with two types of relation extractor demonstrate the superiority of the proposed approach over the state-of-the-art, while further ablation studies verify our intuitions and demonstrate the effectiveness of our proposed two techniques. |
| Researcher Affiliation | Collaboration | 1Zhejiang University 2University of Illinois at Urbana Champaign 3Hikvision Research Institute 4University of Southern California |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper mentions 'https://code.google.com/p/word2vec/' for pre-trained word vectors, but does not provide any link or statement regarding the release of their own source code for the described methodology. |
| Open Datasets | Yes | Following the existing literature (Riedel, Yao, and Mc Callum 2010; Lin et al. 2016; Li et al. 2017; Feng et al. 2018; Liu et al. 2017b), we use the New York Times (NYT) dataset as the training set (Mintz et al. 2009). It uses Freebase (Bollacker et al. 2008) to provide distant supervision on the NYT corpus. |
| Dataset Splits | Yes | For hyper-parameters, we reuse part of them from the previous study (Zeng et al. 2015; Li et al. 2017), and tune the rest part by grid-search with the three-fold cross-validation (on the training set). |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, processor types) used for running its experiments. |
| Software Dependencies | No | The paper mentions using word vectors pre-trained by the Skip-gram algorithm (word2vec) but does not specify version numbers for word2vec or any other software dependencies like programming languages or deep learning frameworks. |
| Experiment Setup | Yes | The final hyper-parameter setting used in our experiments are summarized in Table 2. Table 2: Parameter Name Value Candidate Set sentence embedding dimension 100 {100, 150, 200} batch size 100 {100, 150, 200} superbag size 2 {2, 3, 4, 5} sliding window size 3 reused from previous work word vector dimension 50 position embedding dimension 5 dropout probability 0.5 |