A Joint Training Dual-MRC Framework for Aspect Based Sentiment Analysis
Authors: Yue Mao, Yi Shen, Chao Yu, Longjun Cai13543-13551
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct experiments on these subtasks, and results on several benchmark datasets demonstrate the effectiveness of our proposed framework, which significantly outperforms existing state-of-the-art methods. |
| Researcher Affiliation | Industry | Yue Mao, Yi Shen, Chao Yu, Longjun Cai Alibaba Group, Beijing, China {maoyue.my, sy133447, aiqi.yc, longjun.clj}@alibaba-inc.com |
| Pseudocode | Yes | Algorithm 1: The inference Process for Triple Extraction of the Dual-MRC Framework |
| Open Source Code | No | The paper mentions 'https://github.com/google-research/bert' which is for the BERT model used as a backbone, but there is no explicit statement or link providing the source code for the authors' proposed framework. |
| Open Datasets | Yes | Original datasets are from the Semeval Challenges(Pontiki et al. 2014, 2015, 2016), where ATs and corresponding sentiment polarities are labeled. We evaluate our framework on three public datasets derived from them. The first dataset is from (Wang et al. 2017), where labels for opinion terms are annotated. All datasets share a fixed training/test split. The second dataset is from (Fan et al. 2019), where (AT, OT) pairs are labeled. The third dataset is from (Peng et al. 2020) where (AT, OT, SP) triples are labeled. |
| Dataset Splits | Yes | Also, 20% of the data from the training set are randomly selected as the validation set. |
| Hardware Specification | Yes | All experiments are conducted on a single Tesla-V100 GPU. |
| Software Dependencies | No | The paper mentions using BERT models and Adam optimizer, but it does not specify any software dependencies with explicit version numbers (e.g., Python 3.x, PyTorch 1.x). |
| Experiment Setup | Yes | We use Adam optimizer with a learning rate of 2e 5 and warm up over the first 10% steps to train for 3 epochs. The batch size is 12 and a dropout probability of 0.1 is used. The hyperparameters α, β, γ for the final joint training loss in Equation 14 are not sensitive to results, so we fix them as 1/3 in our experiments. |