FL-MSRE: A Few-Shot Learning based Approach to Multimodal Social Relation Extraction
Authors: Hai Wan, Manrong Zhang, Jianfeng Du, Ziling Huang, Yufei Yang, Jeff Z. Pan13916-13923
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | FL-MSRE is empirically shown to outperform the baseline significantly. |
| Researcher Affiliation | Academia | School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou 510006, P.R.China 2 Guangzhou Key Laboratory of Multilingual Intelligent Processing, Guangdong University of Foreign Studies, Guangzhou 510006, P.R.China 3 School of Informatics, The University of Edinburgh, Edinburgh, UK 4 Pazhou Lab, Guangzhou 510330, P.R.China |
| Pseudocode | No | The paper describes the model architecture and training process in prose and diagrams (Figure 3), but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | 1The code and datasets are available at https://github.com/sysulic/FL-MSRE. |
| Open Datasets | Yes | Since existing SRE datasets (Zhang et al. 2015; Sun, Schiele, and Fritz 2017) only contain unimodal information, we construct three datasets from four classical masterpieces and corresponding TV series... The code and datasets are available at https://github.com/sysulic/FL-MSRE. |
| Dataset Splits | Yes | Following the traditional way of few-shot learning (Han et al. 2018; Gao et al. 2019), we use disjoint sets of social relations for training, validation and test. For DRC-TF, we respectively use 3, 3 and 3 relations for training, validation and test. For OW-TF, we respectively use 5, 5 and 5 relations for training, validation and test. For FC-TF, we respectively use 14, 5 and 5 relations for training, validation and test. |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU or CPU models, or memory amounts used for running experiments. |
| Software Dependencies | No | The paper mentions using Face Net and BERT models, but does not provide specific version numbers for these or other software dependencies like programming languages or libraries. |
| Experiment Setup | Yes | In all of our experiments, we use the Adam optimizer (Kingma and Ba 2015) and tune hyperparameters to the best values according to the accuracy on the validation set. To be specific, the batch size is set to 8 and the weight delay is set to 0.1. The learning rate is initialized as 2 10 5 for the baseline and initialized as 2 10 6 for our approach. |