Distantly Supervised Entity Relation Extraction with Adapted Manual Annotations

Authors: Changzhi Sun, Yuanbin Wu7039-7046

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on the benchmark NYT dataset show that our approach significantly outperforms state-of-the-art methods.
Researcher Affiliation Academia Changzhi Sun,1 Yuanbin Wu1,2 1School of Computer Science and Software Engineering, East China Normal University 2Shanghai Key Laboratory of Multidimensional Information Processing changzhisun@stu.ecnu.edu.cn, ybwu@cs.ecnu.edu.cn
Pseudocode Yes Algorithm 1 The training procedure
Open Source Code No 1We will make our implementation publicly available.
Open Datasets Yes We evaluate the proposed framework on public NYT dataset. 5The training set has 353k relation triples, which are generated by distant supervision. ... 5https://github.com/shanzhenren/Co Type. We use standard ACE05 dataset as the manually labeled dataset (see Figure 1) 6. ... 6https://github.com/tticoin/LSTM-ER.
Dataset Splits Yes Following (Ren et al. 2017; Zheng et al. 2017; Wang et al. 2018), we randomly select 10% of the test set as the development set and use the remaining data as evaluation.
Hardware Specification No The paper does not provide specific hardware details such as CPU/GPU models, memory specifications, or cloud computing instance types used for running the experiments.
Software Dependencies No The paper mentions using 'bi LSTM', 'CNNs', 'Adadelta', 'dropout' and 'glove word embeddings' but does not provide specific version numbers for any software, libraries, or frameworks used.
Experiment Setup Yes Our word embeddings is initialized with 100-dimensional glove (Pennington, Socher, and Manning 2014) word embeddings. The dimensionality of the hidden units is 128. For all CNN in our network, the kernel sizes are 2 and 3, and the output channels are 25. We optimize our model using Adadelta (Zeiler 2012) with gradient clipping. The network is regularized with dropout.