Entity Guided Question Generation with Contextual Structure and Sequence Information Capturing
Authors: Qingbao Huang, Mingyi Fu, Linzhang Mo, Yi Cai, Jingyun Xu, Pijian Li, Qing Li, Ho-fung Leung13064-13072
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Both automatic and manual metrics show that our model can generate comparable questions with state-of-the-art models. Our code is available at https://github.com/VISLANG-Lab/EGSS. |
| Researcher Affiliation | Academia | 1School of Software Engineering, South China University of Technology, Guangzhou, China 2School of Electrical Engineering, Guangxi University, Nanning, China 3Key Laboratory of Big Data and Intelligent Robot (SCUT), MOE of China 4Department of Computing, The Hong Kong Polytechnic University, Hong Kong, China 5Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong, China |
| Pseudocode | No | The paper describes the model architecture and equations but does not include a clearly labeled pseudocode or algorithm block. |
| Open Source Code | Yes | Our code is available at https://github.com/VISLANG-Lab/EGSS. |
| Open Datasets | Yes | We conduct experiments on the accessible part of widely-used SQu AD datasets (Rajpurkar et al. 2016). |
| Dataset Splits | Yes | In Split1, the original dev set is used as test set, and the original training set is randomly divided into training set and dev set at a 9:1 ratio. In Split2, the SQu AD training set remains the same and the original dev set is randomly split into our dev set and test set with the ratio 1:1. |
| Hardware Specification | No | The paper does not specify any hardware details (e.g., GPU models, CPU types, or memory specifications) used for running the experiments. |
| Software Dependencies | No | The paper mentions using "pre-trained Glove word embeddings" and an "SGD optimizer," but does not provide specific version numbers for any software dependencies like libraries, frameworks, or programming languages. |
| Experiment Setup | Yes | We set the dimension of Answer tags, POS tags, and NER tags with 100, 60, and 50, respectively. The hidden size of LSTM are 300 for all encoder and decoder. For the GCN encoder, the level of stacked layer is 4. We make the vocabulary with top 45000 frequency words. When training, we use SGD optimizer with 0.2 dropout rate, 0.01 learning rate and train our model for 30 epochs. For the decoder part, we use beam search with beam size of 12 to get the final result. |