GANs for Semi-Supervised Opinion Spam Detection
Authors: Gray Stanton, Athirai A. Irissappane
IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on Trip Advisor data show that spam GAN outperforms existing techniques when labeled data is limited. We conduct experiments on Trip Advisor dataset and show that spam GAN outperforms existing works when using limited labeled data. |
| Researcher Affiliation | Academia | 1Department of Statistics, Colorado State University 2School of Engineering and Technology, University of Washington, Tacoma gray.stanton@colostate.edu, athirai@uw.edu |
| Pseudocode | Yes | Algorithm 1: spam GAN |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described. |
| Open Datasets | Yes | We use the Trip Advisor labeled dataset [Ott et al., 2011] 5, consisting of 800 truthful reviews on Chicago hotels and 800 deceptive reviews obtained from Amazon Mechanical Turk. We augment the labeled set with 32, 297 unlabeled Trip Advisor reviews for Chicago hotels 6. 5http://myleott.com/op-spam.html 6http://times.cs.uiuc.edu/ wang292/Data/index.html |
| Dataset Splits | No | We use a 80/20 train-test split on labeled data. The paper does not explicitly mention a separate validation set split or how hyperparameter tuning was performed if not using a validation set, nor does it specify cross-validation. |
| Hardware Specification | Yes | The train time of spam GAN using a Tesla P4 GPU was 1.5 hrs. |
| Software Dependencies | No | The paper mentions ADAM optimizer, GRU layers, and variational dropout, but does not provide specific version numbers for these or any other software components. |
| Experiment Setup | Yes | In spam GAN, the generator consists of 2 GRU layers of 1024 units each...word embeddings with dimension 50. For generator, learning rate = 0.001, weight decay =1 10 7. Gradient clipping is set to a maximum global norm of 5. The discriminator contains 2 GRU layers of 512 units each...Learning rate =0.0001 and weight decay =1 10 4. We set balancing coefficient β = 1. |