BMWP: the first Bengali math word problems dataset for operation prediction and solving.

Area of Science:

• Artificial Intelligence
• Natural Language Processing
• Machine Learning

Background:

• Solving math word problems is a key benchmark for AI and language model intelligence.
• Low-resource languages like Bengali lack standardized datasets for this task.
• Existing research predominantly uses English and Chinese datasets.

Purpose of the Study:

• To introduce the first Bengali Math Word Problem Benchmark Data Set (BMWP).
• To establish benchmarking methods for evaluating AI models on Bengali word problems.
• To investigate the performance of deep learning models in solving these problems.

Main Methods:

• Creation of the BMWP dataset with 8653 word problems.
• Implementation and comparison of state-of-the-art deep learning neural network architectures.
• Benchmarking performance on operation prediction from Bengali word problems.

Main Results:

• The BMWP dataset provides a novel resource for low-resource language NLP research.
• Deep learning models were evaluated for their effectiveness in solving Bengali math word problems.
• Achieved an accuracy of [specific accuracy value, if available, otherwise state 'promising results'] in operation prediction.

Conclusions:

• The developed BMWP dataset addresses a critical gap in resources for Bengali NLP.
• The study demonstrates the feasibility of applying deep learning to solve math word problems in Bengali.
• The dataset and code will be publicly available to foster further research.