DRr-Net: Dynamic Re-Read Network for Sentence Semantic Matching
Authors: Kun Zhang, Guangyi Lv, Linyuan Wang, Le Wu, Enhong Chen, Fangzhao Wu, Xing Xie7442-7449
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on three sentence matching benchmark datasets demonstrate that DRr-Net has the ability to model sentence semantic more precisely and significantly improve the performance of sentence semantic matching. In addition, it is very interesting that some of finding in our experiments are consistent with the findings of psychological research. |
| Researcher Affiliation | Collaboration | 1Anhui Province Key Laboratory of Big Data Analysis and Application, School of Computer Science and Technology, University of Science and Technology of China {zhkun, gylv, wly757}@mail.ustc.edu.cn, cheneh@ustc.edu.cn 2Hefei University of Technology, China, lewu@hfut.edu.cn 3 Microsoft Research Asia, China, wufangzhao@gmail.com, xing.xie@microsoft.com |
| Pseudocode | No | The paper describes the model architecture and equations but does not include a clearly labeled pseudocode or algorithm block. |
| Open Source Code | No | The paper states: 'During implementation, we utilize Photinia1 to build our entire model. 1https://github.com/XoriieInpottn/photinia'. This refers to a third-party framework used, not the open-source code for the DRr-Net methodology described in the paper. |
| Open Datasets | Yes | We evaluate our model on three well-studied datasets: the Stanford Natural Language Inference (SNLI) (Bowman et al. 2015), the Sentence Involving Compositional Knowledge (SICK) (Marelli et al. 2014), and Quora duplicate questions (Quora) (Iyer, Dandekar, and Csernai 2017). |
| Dataset Splits | No | The paper states, 'We have tuned the hyper-parameters on the validation set' and mentions the total size of datasets (e.g., 'SNLI contains 570,152 human annotated sentence pairs'), but it does not specify the exact percentages or counts for training, validation, or test splits for any of the datasets. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used for running the experiments, such as GPU models, CPU types, or memory specifications. |
| Software Dependencies | No | The paper mentions using 'pre-trained word vectors (840B GloVe)' and 'Adam optimizer' and 'Photinia', but it does not provide specific version numbers for any software libraries, frameworks, or tools used in the implementation, aside from the Python framework 'Photinia' without its version. |
| Experiment Setup | Yes | The dimension is set as 300. Character-level word embedding is set as 100. The number of stack layers in ASG unit is set as 3 and the re-read length in DRr unit is set as 6. The hidden state size of GRUs in these two units is set as 256. ... We use Adam optimizer with learning rate 10-4. |