Generating Chinese Ci with Designated Metrical Structure
Authors: Richong Zhang, Xinyu Liu, Xinwei Chen, Zhiyuan Hu, Zhaoqing Xu, Yongyi Mao7459-7467
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We now study the performance of the proposed MRCG model. Ideally we would like to compare MRCG with some recent art in Ci generation. ... Table 1 shows the metrical performances of the compared models under the three metrics. |
| Researcher Affiliation | Academia | Richong Zhang,1,2 Xinyu Liu,3 Xinwei Chen,3 Zhiyuan Hu,4 Zhaoqing Xu,2 Yongyi Mao 3 1SKLSDE, School of Computer Science and Engineering, Beihang University, Beijing, China 2Beijing Advanced Institution on Big Data and Brain Computing, Beihang University, Beijing, China 3School of Electrical Engineering and Computer Science, University of Ottawa, Ottawa, Ontario 4Beijing University of Chemical Technology, Beijing, China |
| Pseudocode | No | The paper describes the model architecture (encoder, decoder) with equations and diagrams, but does not include a formal pseudocode or algorithm block. |
| Open Source Code | No | It is unfortunate however that the papers presenting those models do not include sufficient implementation details, nor did the authors make available their source code. |
| Open Datasets | Yes | We extract a dataset from a Chinese poetry website(Yun 2017). The dataset contains 82,724 Ci s, written for 818 Cipai s. |
| Dataset Splits | No | The dataset is split into the training set and the testing set as follows. For each Cipai, we randomly select 5% of its Ci s to assemble the testing set. ... In total, we have 3,797 Ci s in the testing set and 78,927 in the training set. |
| Hardware Specification | No | No specific hardware details (like GPU/CPU models, memory, or cloud instances) are mentioned for running experiments. |
| Software Dependencies | No | Both the Seq2Seq model and the attention model are built using the tensorflow NMT tool (Luong, Brevdo, and Zhao 2017). |
| Experiment Setup | Yes | In our implementation of MRCG, we choose the latent semantic dimension K = 64. Other dimensions are chosen as Kw = Kw = Ks = 128, and Ku = 256. |