Datasets:
license: cc-by-4.0
language:
- sq
- as
- en
- zh
- hi
- si
- ta
- ur
- or
size_categories:
- 10K<n<100K
MultiMWP: A Multi-Way Parallel Dataset for Math Word Problem Generation
Overview
MultiMWP is a multi-way parallel dataset designed for math word problem (MWP) generation across 9 languages. The dataset consists of structured math word problems in plain text format. It is intended for problem generation rather than problem-solving. The same dataset is in Github: https://github.com/OmegaGamage/multiMWP/tree/master/dataset
Dataset Structure
MultiMWP includes two categories of math word problems:
- Simple Category
- Algebraic Category
Each category is provided as a plain text file per language.
Languages
The dataset is available in the following 9 languages:
- Albanian (
sqi) - Assamese (
asm) - Chinese (
zho) - English (
eng) - Hindi (
hi) - Oriya (
ory) - Sinhala (
sin) - Tamil (
tam) - Urdu (
urd)
Data Format
Each file contains a list of math word problems, one per line. The dataset does not include solutions, equations, or additional metadata.
Example from Simple-English.txt:
Bill has 9 marbles and Jim has 7 fewer marbles than Bill. How many marbles does Jim have?
Amal has 10 apples and Vimal has 8 fewer apples than Amal. How many apples does Vimal have?
Kamal has 16 marbles and Nimal has 12 fewer marbles than Kamal. How many marbles does Nimal have?
Example from Algebraic-English.txt:
Find three consecutive odd integers such that the sum of the first integer twice the second integer and three times the third is 70.
The sum of two numbers is 120, if the larger number is 4 times the smaller number, find the two numbers?
A number is twice another number, if their sum is 48 find the numbers.
Dataset Size
The dataset consists of:
- Simple Category: 3160 MWPs × 9 languages
- Algebraic Category: 4210 MWPs × 9 languages
License
This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) License. This means you are free to share, modify, and use this dataset for any purpose, as long as you provide appropriate credit. For details, see: CC BY 4.0 License.
Citation
If you use this dataset in your research, please cite:
Gamage, Omega, et al. "A Multilingual Dataset (MultiMWP) and Benchmark for Math Word Problem Generation." IEEE Transactions on Audio, Speech and Language Processing (2025).
Bibtex
@ARTICLE{10933586,
author={Gamage, Omega and Ranathunga, Surangika and Lee, Annie and Sun, Xiao and Singh, Aryaveer and Skenduli, Marjana Prifti and Alam, Mehreen and Nayak, Ajit Kumar and Gao, Haonan and Deori, Barga and Ji, Jingwen and Zhang, Qiyue and Zeng, Yuchen and Tian, Muxin and Mao, Yanke and Trico, Endi and Nako, Danja and Shqezi, Sonila and Hoxha, Sara and Imami, Dezi and Doksani, Dea and Pandey, Virat Kumar and Ananya, Ananya and Aggarwal, Nitisha and Hussain, Naiyarah and Dwivedi, Vandana and Sinha, Rajkumari Monimala and Kalita, Dhrubajyoti},
journal={IEEE Transactions on Audio, Speech and Language Processing},
title={A Multilingual Dataset (MultiMWP) and Benchmark for Math Word Problem Generation},
year={2025},
volume={},
number={},
pages={1-13},
keywords={Translation;Multilingual;Mathematical models;Arithmetic;Natural language processing;Speech processing;Data models;Data mining;Artificial neural networks;Training;Benchmark;low-resource languages;math word problem generation;multi-way parallel dataset;multilingual dataset},
doi={10.1109/TASLPRO.2025.3552936}
}