Tensor Graph Convolutional Networks for Text Classification
Authors: Xien Liu, Xinxin You, Xiao Zhang, Ji Wu, Ping Lv8409-8416
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments are conducted on benchmark datasets, and the results illustrate the effectiveness of our proposed framework. Experiments and results analysis In this section, we evaluate the performance of our proposed Tensor GCN based text classification framework, then carefully examine the effectiveness of our constructed text graph tensor and the ability of our developed Tensor GCN algorithm for joint learning on multi-graphs. |
| Researcher Affiliation | Collaboration | 1Department of Electronic Engineering, Tsinghua University, Beijing 100084, China 2Institute for Precision Medicine, Tsinghua University, Beijing 100084, China 3Tsinghua-i FLYTEK Joint Lab, i Flytek Research, Beijing 100084, China 4State Key Laboratory of Cognitive Intelligence, Hefei, Anhui 230088, China |
| Pseudocode | No | The paper describes methods and processes in text and with mathematical formulas, but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to source code, nor does it explicitly state that code will be released or is available. |
| Open Datasets | Yes | A suite of recently widely used benchmark datasets were used to perform experiments and analysis. The benchmark corpora consist of five text classification datasets: 20Newsgroups dataset, Ohsumed dataset, R52 Reuters dataset, R8 Reuters dataset, and Movie Review dataset. ... More detail descriptions about baselines and datasets can be found in (Yao, Mao, and Luo 2019). |
| Dataset Splits | Yes | As used in (Yao, Mao, and Luo 2019), we randomly selected 10% of the training set as the validation set, which labels are not be used for training. |
| Hardware Specification | No | The paper does not provide specific hardware details such as exact GPU/CPU models or processor types used for running its experiments. |
| Software Dependencies | No | The paper mentions 'Stanford Core NLP parser' and 'Glove' but does not provide specific version numbers for any software dependencies or libraries. |
| Experiment Setup | Yes | In the construction of the sequential-based graph, the window size is 20 as used in (Yao, Mao, and Luo 2019), and in the semantic-based graph construction, the initial word embeddings are pre-trained with Glove, and the dimension is 300, the dimension of LSTM is also set to 300. As suggested and used in (Yao, Mao, and Luo 2019), in this study, we also take a two-layers Tensor GCN and dimension of the node embedding in the first layer is 200, and the dimension of the node embedding in the second layer is equal to the number of the labels. In the training process, the dropout rate is 0.5, and L2 loss weight is 5e-6. As used in (Yao, Mao, and Luo 2019), we randomly selected 10% of the training set as the validation set, which labels are not be used for training. A maximum of 1000 epochs and Adam optimizer with a learning rate of 0.002 are used, and early stopping is performed when validation loss does not decrease for ten consecutive epochs. All results reported in this study are the mean values of ten independent runs. |