Rich Event Modeling for Script Event Prediction
Authors: Long Bai, Saiping Guan, Zixuan Li, Jiafeng Guo, Xiaolong Jin, Xueqi Cheng
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on the widely used Gigaword Corpus show the effectiveness of the proposed framework. We conduct extensive experiments on the widely used Gigaword corpus, which show the superiority of the proposed framework. |
| Researcher Affiliation | Academia | Long Bai1,2, Saiping Guan1,2, Zixuan Li1,2, Jiafeng Guo1,2, Xiaolong Jin1,2, Xueqi Cheng1,2 1 CAS Key Lab of Network Data Science and Technology, Institute of Computing Technology, Chinese Academy of Sciences (CAS) 2 School of Computer Science and Technology, University of Chinese Academy of Sciences |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide an explicit statement or link for the open-source code of the proposed REP framework. |
| Open Datasets | Yes | We use two datasets to evaluate the proposed framework... MCNC dataset (Granroth-Wilding and Clark 2016) is extracted from the New York Time portion of the Gigaword corpus (Graff et al. 2003)... MCNC-rich dataset is proposed in this paper... It is extracted from the same corpus with MCNC. |
| Dataset Splits | Yes | Basic statistics of the two datasets are shown in Table 2. MCNC: # Train Instances 1,440,295 # Dev Instances 10,000 # Test Instances 10,000. MCNC-rich: # Train Instances 1,006,301 # Dev Instances 10,000 # Test Instances 10,000. |
| Hardware Specification | Yes | All the experiments are conducted on Tesla V100. |
| Software Dependencies | No | The paper mentions using SPRING parser and Allen NLP but does not provide specific version numbers for these or other ancillary software dependencies. |
| Experiment Setup | Yes | The length of the narrative event chain Ne is set to 8; the number of the candidate events Nc is set to 5; word embedding dimension dw is set to 300; event embedding dimension de is set to 128; the number of layers for rich event encoder is selected from {1, 2}; the dimension of feedforward network in rich event encoder is selected from {512, 1024}; the number of heads for rich event encoder is set to 8; the number of layers for temporal integration is set to 2; the dimension of feedforward network in temporal integration is set to 1024; the number of heads for temporal integration is set to 16; the dropout rate is set to 0.1; the learning rate is set to 1e-3; the regularization factor λ is set to 1e-5; The best settings (underlined) are selected according to the performance on development set. |