Data and code of the Findings of EMNLP'23 paper "MuG: A Multimodal Classification Benchmark on Game Data with Tabular, Textual, and Visual Fields". For any suggestion/question, please feel free to create an issue or drop an email @ (jiaying.lu@emory.edu).
Table of Contents
The eight datasets used in paper can be downloaded from https://doi.org/10.6084/m9.figshare.21454413.
After downloading and decompressing them under ./datasets/
directory, the directory looks like:
š ./dataset
|-- š Pokemon-primary_type
|-- š Pokemon-secondary_type
|-- š Hearthstone-All-cardClass
|-- š Hearthstone-All-set
|-- š Hearthstone-Minion-race
|-- š Hearthstone-Spell-spellSchool
|-- š LeagueOfLegends-Skin-category
|-- š CSGO-Skin-quality
|-- CHANGELOG
And each subdirectory represents one dataset, for instance:
š ./dataset/Pokemon-primary_type
|-- info.json
|-- train.csv
|-- dev.csv
|-- test.csv
|-- train_images.zip
|-- dev_images.zip
|-- test_images.zip
where info.json
stores meta information of the dataset;
train/dev/test.csv
store raw tabular and text features of each sample;
train_images/dev_images/test_images.zip
represent a compressed directory of raw images.
All dependecies are listed in conda_env.yml. We recommend using conda
to manage the environment.
conda env create -n MuG_env --file conda_env.yml
Example scripts to run unimodal classifiers and multimodal classifiers are listed in run_baseline.sh.
For instance, we can use the following script to reproduce the proposed MuGNet model:
# Run MuGNet modal
python -m baselines.MuGNet.exec \
--dataset_dir datasets/Pokemon_PrimaryType \
--exp_save_dir exps/Pokemon_PrimaryType_mugnet \
--fit_time_limit 28800
where --dataset_dir
specifies the dataset directory, --exp_save_dir
specifies the destination to store the final output, --fit_time_limit
specifies the time limit for the model to run in seconds.
Another example to run GBM model:
python -m baselines.autogluon.exec \
--dataset_dir datasets/Pokemon_PrimaryType \
--exp_save_dir exps/Pokemon_PrimaryType_GBMLarge \
--fit_time_limit 28800 \
--fit_setting GBMLarge
@inproceedings{lu2023MuG,
title = {MuG: A Multimodal Classification Benchmark on Game Data with Tabular, Textual, and Visual Fields},
author = {Jiaying Lu and Yongchen Qian and Shifan Zhao and Yuanzhe Xi and Carl Yang},
doi={10.18653/v1/2023.findings-emnlp.354},
Series = {Findings-EMNLP'23},
booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2023},
month = {December},
year = {2023}
}