Scholarlink

In Mathematical word problems (MWPs) are challenging due to semantic gap between natural language texts and their corresponding mathematical equations. The MWP task aims to generate an appropriate equation as output to solve a given written math problem. However, the performance of such systems may be limited by linguistic characteristics of target language. This paper introduces a model developed to automatically solve Turkish MWPs based on semantic information in input text. The model consists of a bidirectional encoder and a decoder with an attention mechanism. Despite the grammatical complexity and agglutinative structure of the Turkish language, the proposed model achieved an accuracy of 72% on evaluated datasets.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

GRU–GNN Hybrid Approach for Turkish Mathematical Word Problem Solving

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Taha Kubilay Şener,
  • İlhan Aydın

摘要

In Mathematical word problems (MWPs) are challenging due to semantic gap between natural language texts and their corresponding mathematical equations. The MWP task aims to generate an appropriate equation as output to solve a given written math problem. However, the performance of such systems may be limited by linguistic characteristics of target language. This paper introduces a model developed to automatically solve Turkish MWPs based on semantic information in input text. The model consists of a bidirectional encoder and a decoder with an attention mechanism. Despite the grammatical complexity and agglutinative structure of the Turkish language, the proposed model achieved an accuracy of 72% on evaluated datasets.