Event Process Typing via Hierarchical Optimal Transport

Authors: Bo Zhou, Yubo Chen, Kang Liu, Jun Zhao

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that our model outperforms the baseline models, illustrating the effectiveness of our model.
Researcher Affiliation Academia 1 School of Artificial Intelligence, University of Chinese Academy of Sciences 2 National Laboratory of Pattern Recognition, CASIA 3 Beijing Academy of Artificial Intelligence {bo.zhou,yubo.chen,kliu,jzhao}@nlpr.ia.ac.cn
Pseudocode Yes Algorithm 1: HOT algorithm
Open Source Code No The paper does not provide a direct link to its source code or state that the code for their methodology is publicly available.
Open Datasets Yes We use dataset released by Chen et al. (2020), in which each event process with one action and one object label is extracted from wiki How1. It s an online wikistyle publication featuring expert co-authored, how-to articles on a variety of topics and each article consists of a title of how to do something and a series of sentences describing the steps.
Dataset Splits Yes The split of the training/dev./test set is 80/10/10%...
Hardware Specification No The paper mentions using "RoBERTa-base pretrained model from Hugging-Face s Transformers library" but does not specify any hardware details (e.g., GPU model, CPU, memory) used for training or inference.
Software Dependencies No The paper mentions using "RoBERTa-base pretrained model from Hugging-Face s Transformers library (Wolf et al. 2020)", "Word Net (Miller 1995)", "Allen NLP (Gardner et al. 2018)" and "Adam (Kingma and Ba 2015)". However, it does not provide specific version numbers for the Transformers library, Allen NLP, or any other software components.
Experiment Setup Yes The cosine function is chosen as the cost function in the hierarchical optimal transport and the hyper-parameter α is set to 0.1. Adam (Kingma and Ba 2015) is used for optimization with an initial learning rate 1e-4. The models are trained for 50 epochs with batch size of 64.