Schema-Guided Multi-Domain Dialogue State Tracking with Graph Attention Neural Networks
Authors: Lu Chen, Boer Lv, Chi Wang, Su Zhu, Bowen Tan, Kai Yu7521-7528
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiment results show that our approach obtains new state-of-the-art performance on both Multi WOZ 2.0 and Multi WOZ 2.1 benchmarks. |
| Researcher Affiliation | Academia | Mo E Key Lab of Artificial Intelligence Speech Lab, Department of Computer Science and Engineering Shanghai Jiao Tong University, Shanghai, China {chenlusz, boerlv, wangchi16, paul2204, tanbowen, kai.yu}@sjtu.edu.cn |
| Pseudocode | No | The paper describes methods using mathematical equations and textual explanations but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper mentions "The official code https://github.com/jasonwu0731/trade-dst is used." but this refers to a baseline model (TRADE), not the authors' own proposed SST model. No statement or link regarding the open-sourcing of their own code was found. |
| Open Datasets | Yes | We choose Multi WOZ 2.0 (Budzianowski et al. 2018) and Multi WOZ 2.1 (Eric et al. 2019) as our training and evaluation datasets |
| Dataset Splits | No | The paper mentions "It contains 55718 turn-level training samples from 8133 dialogues (only five domains) with 7368 turns of the test set." but does not explicitly specify the size or percentage of a validation set, nor explicit percentages for the splits (e.g., 80/10/10). |
| Hardware Specification | No | No specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments were mentioned in the paper. |
| Software Dependencies | No | The paper mentions using GloVe embeddings and RMSProp for training, but does not specify any software names with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | The number of GAMT steps/layers L1 is three on both datasets. Each model is trained by RMSProp with a learning rate of 0.0001 and a batch size of 32. |