Skip to content

Latest commit

 

History

History
41 lines (27 loc) · 1.31 KB

README.md

File metadata and controls

41 lines (27 loc) · 1.31 KB
Raw

SCA-GPS

Introduction

Code for paper A Symbolic Characters Aware Model for Solving Geometry Problems - ACM MM 2023

Run

Environment

transformers==4.17.0

allennlp==0.9.0

Refined GeoQA Dataset

Please note, we refined the GeoQA dataset to remove the Alpha Chanel in the geometry diagrams to satisfy the requirement of ViT input. The refined dataset named as GeoQA-Pro and used in this repo.

Prepare Roberta-CHN

Due to LFS space limited, please download roberta-chn from https://huggingface.co/hfl/chinese-roberta-wwm-ext and move the pytorch_model.bin to the roberta folder in this repo.

Training

allennlp train config/DPE.json --include-package DPE -s test/

Evaluating

allennlp evaluate test/  GeoQA-Data/Geo-Pro/pro_test.pk --include-package DPE-test --cuda-device 0

Citation

If the paper or the code helps you, please cite the paper in the following format :

@inproceedings{ning2023SCAGPS, 
    author = {Ning, Maizhen and Wang, Qiu-Feng and Huang, Kaizhu and Huang, Xiaowei}, 
    title = {A Symbolic Characters Aware Model for Solving Geometry Problems}, 
    year = {2023}, 
    doi = {10.1145/3581783.3612570}, 
    booktitle = {Proceedings of the 31st ACM International Conference on Multimedia}, 
    series = {MM '23} 
}