DPLM: A Diachronic Language Model for Long-Time Span Classical Chinese

This repository provides a reference implementation of the following paper:

A Diachronic Language Model for Long-Time Span Classical Chinese
Yuting Wei, Meiling Li, Yangfu Zhu, Yuanxing Xu, Yuqing Li, Bin Wu
Information Processing and Management

Prerequisites

To set up the environment, follow these steps:

1. Create an Anaconda environment with Python 3.8:

   conda create -n temporal python=3.8
   conda activate temporal

2. Install the required PyTorch version and dependencies:

   conda install pytorch==1.10.1 torchvision==0.11.2 torchaudio==0.10.1 cudatoolkit=11.3 -c pytorch -c conda-forge

3. Install the remaining Python packages:

   pip install -i https://pypi.tuna.tsinghua.edu.cn/simple -r requirements.txt

4. Download our training dataset:

   Please download the dataset from Google Drive here and place it in the train_datasets folder.

Training

To train the model, simply run the following command:

bash run.sh

Evaluation

To evaluate the model, use the following command, replacing path_to_trained_dplm with the path to your trained model:

python evaluate.py --model_name path_to_trained_dplm

Citation

Please cite our paper if you use our code, dataset, or compare with our model:

@article{WEI2025103925,
title = {A diachronic language model for long-time span classical Chinese},
journal = {Information Processing & Management},
volume = {62},
number = {1},
pages = {103925},
year = {2025},
issn = {0306-4573},
doi = {https://doi.org/10.1016/j.ipm.2024.103925},
url = {https://www.sciencedirect.com/science/article/pii/S030645732400284X},
author = {Yuting Wei and Meiling Li and Yangfu Zhu and Yuanxing Xu and Yuqing Li and Bin Wu}
}

Acknowledgements

The code structure is inspired by TempoBERT. The eval datasets are adapted from the Ancient Chinese Corpus with Word Sense Annotation. We sincerely thank the authors of these projects for their contributions to the community.