Counterfactual-Enhanced Information Bottleneck for Aspect-Based Sentiment Analysis

Authors: Mingshan Chang, Min Yang, Qingshan Jiang, Ruifeng Xu

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on five benchmark ABSA datasets show that our CEIB approach achieves superior prediction performance and robustness over the state-of-the-art baselines.
Researcher Affiliation Academia 1Shenzhen Key Laboratory for High Performance Data Mining, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences 2University of Chinese Academy of Sciences 3Harbin Institute of Technology (Shenzhen)
Pseudocode No The paper describes the methodology using text and mathematical equations but does not include a structured pseudocode or algorithm block.
Open Source Code Yes Code and data to reproduce the results in this paper is available at: https://github.com/shesshan/CEIB.
Open Datasets Yes We conduct our experiments on five benchmark ABSA datasets: REST14 and LAP14 from (Pontiki et al. 2014), REST15 from (Pontiki et al. 2015), REST16 from (Pontiki et al. 2016), and MAMS from (Jiang et al. 2019). We also use ARTS dataset (Xing et al. 2020), including REST14-ARTS and LAP14-ARTS, to test the robustness of the ABSA models.
Dataset Splits Yes We adopt the official data splits, which keep the same as in the original papers. The detailed statistics of the utilized datasets are shown in Table 1, which includes Train, Dev (for MAMS), and Test splits.
Hardware Specification No The paper does not provide specific hardware details such as GPU models, CPU types, or cloud computing instance specifications used for experiments.
Software Dependencies No The paper mentions software like PyTorch, T5-XXL, and BERT, but does not provide specific version numbers for these software components or libraries.
Experiment Setup Yes We train all our models for 30 epochs. Adam is used as the optimizer with the initial learning rate as 5e 5 and the weight decay as 1e 4. The hyper-parameter α is set in range 0.5 to 1.0 and β for l2-norm regularization is adaptively set by the optimizer.