Character-Based Parsing with Convolutional Neural Network

Authors: Xiaoqing Zheng, Haoyuan Peng, Yi Chen, Pengjing Zhang, Wenqiang Zhang

IJCAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that the KMCNN can successfully recover the structure of Chinese sentences. We conducted two sets of experiments.
Researcher Affiliation Academia Xiaoqing Zheng, Haoyuan Peng, Yi Chen, Pengjing Zhang, Wenqiang Zhang School of Computer Science, Fudan University, Shanghai, China Shanghai Key Laboratory of Intelligent Information Processing {zhengxq, penghy11, yichen11, pengjingzhang13, wqzhang}@fudan.edu.cn
Pseudocode No The paper describes the dynamic programming decoding algorithm in text and refers to a figure, but it does not provide explicit pseudocode or algorithm blocks.
Open Source Code No The paper mentions using a third-party tool 'Word2Vec' and provides a link for it, but it does not state that the code for their own described methodology (KMCNN) is open-source or provide a link to it.
Open Datasets Yes We compared the performance of the KMCNN with the existing stateof-the-art systems on the Penn Chinese Treebank 5 (CTB-5) using the standard split of data1. We transformed the parse trees of the CTB-5 into their binary forms for training by the method described in Section 2.
Dataset Splits Yes Sections 1-270, 400-931, 1001-1151 are used for training, sections 301-325 for development, and sections 271-300 for testing.
Hardware Specification No The paper does not provide any specific details about the hardware used for running its experiments.
Software Dependencies No The paper mentions the 'Word2Vec tool' but does not specify any software dependencies with version numbers for its own implementation or experimental setup.
Experiment Setup Yes Table 1: Hyper-parameters of the network. Hyper-parameter Value Window sizes of two convolutional layers 5/3 Values k in two k-max pooling layers 5/3 Number of convolutional units 200 Number of hidden units 300 Character feature dimension 50 Learning rate 0.02 Penalization term κ 0.05 Regularization parameter λ 0.0001