Time to Transfer: Predicting and Evaluating Machine-Human Chatting Handoff
Authors: Jiawei Liu, Zhe Gao, Yangyang Kang, Zhuoren Jiang, Guoxiu He, Changlong Sun, Xiaozhong Liu, Wei Lu5841-5849
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To provide insights into the task and validate the proposed model, we collect two new datasets. Extensive experimental results are presented and contrasted against a series of baseline models to demonstrate the efficacy of our model on MHCH. |
| Researcher Affiliation | Collaboration | Jiawei Liu,1 Zhe Gao,2 Yangyang Kang,2 Zhuoren Jiang,3 Guoxiu He,1 Changlong Sun,2,3 Xiaozhong Liu,4 Wei Lu1 1Wuhan University, Wuhan, China 2Alibaba Group, China 3Zhejiang University, Hangzhou, China 4Indiana University Bloomington, Bloomington, United States |
| Pseudocode | No | The paper describes the model architecture and its components in text and with a diagram, but it does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | (3) To assist other scholars in reproducing the experiment outcomes and further investigating this novel but important problem, two real-world customer service dialogue datasets are collected, labeled2. 2https://github.com/Weijia Lau/MHCH-DAMI |
| Open Datasets | Yes | To the best of our knowledge, there is no publicly available dataset for the MHCH task. To address this problem, we propose two Chinese sales customer service dialogue datasets, namely Clothing and Makeup, which are collected from Taobao4, one of the largest decentralized E-commerce platforms in the world... 2https://github.com/Weijia Lau/MHCH-DAMI |
| Dataset Splits | Yes | After preprocessing, the datasets are partitioned for training, validation and test with an 80/10/10 split. |
| Hardware Specification | Yes | All the methods are implemented by Tensorflow6 and run on a server configured with a Tesla V100 GPU, 8 CPU, and 16G memory. |
| Software Dependencies | No | All the methods are implemented by Tensorflow6. Footnote 6 points to its official website. Other tools like 'jieba' and 'Snow NLP' are mentioned with links but without specific version numbers. |
| Experiment Setup | Yes | The sizes of LSTM hidden state k, attention units z are all 128, and batch size is set as 128. The dropout (Srivastava et al. 2014) rate is 0.25 and the number of epochs is set as 30. The L2 regularization weight is 10 4. Finally, we train the models with an initial learning rate of 0.0075. |