MCL: Multi-Granularity Contrastive Learning Framework for Chinese NER

Authors: Shan Zhao, ChengYu Wang, Minghao Hu, Tianwei Yan, Meng Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on four Chinese NER datasets show that MCL obtains state-of-the-art results while considering model efficiency.
Researcher Affiliation Academia 1 School of Computer Science and Information Engineering, He Fei University of Technology, He Fei, China 2 College of Computer, National University of Defense Technology, Changsha, China 3 Information Research Center of Military Science, PLA Academy of Military Science, Beijing, China
Pseudocode No The paper does not contain any pseudocode or clearly labeled algorithm blocks.
Open Source Code Yes The source code of the proposed method is publicly available at https://github.com/zs50910/MCL
Open Datasets Yes To evaluate the performance of our method, we conduct experiments on four datasets, including Onto Notes 4.0 (Weischedel et al. 2011), Weibo (Peng and Dredze 2015), MSRA (Levow 2006), and Chinese Resume dataset (Zhang and Yang 2018).
Dataset Splits No The paper mentions using a 'development set' for tuning, but does not provide specific percentages or sample counts for training, validation, or test splits needed for reproduction.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments, such as GPU or CPU models.
Software Dependencies No The paper mentions general libraries like Bi LSTM and CRF, but does not provide specific version numbers for software dependencies such as programming languages, deep learning frameworks, or libraries.
Experiment Setup Yes We regularize our network using dropout with a rate tuned on the development set (the dropout rate is 0.5 for embeddings and encoder). We utilize 1 layer encoder in our network and set the dimensionality of hidden size was set to 100 for Weibo and 300 for the rest three datasets. The learning rate was set to 0.007 for all datasets with Adamax. The temperatures are 0.3 for CCL and 0.05 for BCL.