Leveraging Title-Abstract Attentive Semantics for Paper Recommendation
Authors: Guibing Guo, Bowei Chen, Xiaoyan Zhang, Zhirong Liu, Zhenhua Dong, Xiuqiang He67-74
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments on two real datasets, and show that our approach outperforms other state-of-the-art approaches in terms of accuracy. |
| Researcher Affiliation | Collaboration | 1Northeastern University, China, 2Shenzhen University, China, 3Noah s Ark Research Lab, Huawei, China |
| Pseudocode | No | No pseudocode or clearly labeled algorithm blocks are present in the paper. |
| Open Source Code | No | The paper does not provide any statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | Two real-world datasets are adopted in our experiments, namely citeulike-a and PRSDataset. Both datasets contain the information of title and abstract for papers, and the interactions between users and papers. For each dataset, we randomly split it into two subsets, where 80% of userpaper interactions are classified as the training set and the rest 20% as the testing set. citeulike-a is extracted from Cite ULike3 [http://www.citeulike.org/faq/data.adp], and PRSDataset comes from CSPub Guru4 [https://sites.google.com/site/tinhuynhuit/dataset]. |
| Dataset Splits | No | The paper states a split of '80%... training set and the rest 20% as the testing set' but does not specify a separate validation dataset split. |
| Hardware Specification | No | The paper does not explicitly describe the specific hardware used for running the experiments (e.g., GPU/CPU models, memory specifications). |
| Software Dependencies | No | The paper does not provide specific version numbers for any software components or libraries used in the experiments. It mentions 'Adam gradient descent' but not a version. |
| Experiment Setup | Yes | We have tested the neural batch size in {128, 256, 512, 1024}, the L2 loss weight in {0.1, 0.01, 0.001}, and tuned the semantic weight parameters α and β from 0 to 1 stepping by 0.2 and the learning rate in {0.1, 0.01, 0.001, 0.0001}. Furthermore, we test the number of neurons in each hidden layer and that of latent memory keys from 10 to 100 stepping by 10. All the network parameters are initialized by normal distribution (0, 0.1). We optimize our models with the Adam gradient descent. |