Abstract:Developing automatic Math Word Problem (MWP) solvers is a challenging task that demands the ability of understanding and mathematical reasoning over the natural language. Recent neural-based approaches mainly encode the problem text using a language model and decode a mathematical expression over quantities and operators iteratively. Note the problem text of a MWP consists of a context part and a question part, a recent work finds these neural solvers may only perform shallow pattern matching between the context text and the golden expression, where question text is not well used. Meanwhile, existing decoding processes fail to enforce the mathematical laws into the design, where the representations for mathematical equivalent expressions are different. To address these two issues, we propose a new encoder-decoder architecture that fully leverages the question text and preserves step-wise commutative law. Besides generating quantity embeddings, our encoder further encodes the question text and uses it to guide the decoding process. At each step, our decoder uses Deep Sets to compute expression representations so that these embeddings are invariant under any permutation of quantities. Experiments on four established benchmarks demonstrate that our framework outperforms state-of-the-art neural MWP solvers, showing the effectiveness of our techniques. We also conduct a detailed analysis of the results to show the limitations of our approach and further discuss the potential future work. Code is available at https://github.com/sophistz/Question-Aware-Deductive-MWP.

Tree-structured Decoding for Solving MathWord Problems

A Goal-Driven Tree-Structured Neural Model for Math Word Problems

Graph-To-Tree Learning For Solving Math Word Problems

Deep Learning in Automatic Math Word Problem Solvers

MathDQN: Solving Arithmetic Word Problems Via Deep Reinforcement Learning.

An Expression Tree Decoding Strategy for Mathematical Equation Generation

Structure-Unified M-Tree Coding Solver for Math Word Problem

Specialized Mathematical Solving by a Step-By-Step Expression Chain Generation

Template-Based Math Word Problem Solvers with Recursive Neural Networks

An Edge-Enhanced Hierarchical Graph-to-Tree Network for Math Word Problem Solving

Tackling Math Word Problems with Fine-to-Coarse Abstracting and Reasoning

Graph-to-Tree Neural Networks for Learning Structured Input-Output Translation with Applications to Semantic Parsing and Math Word Problem.

Techniques to Improve Neural Math Word Problem Solvers

Non-Autoregressive Math Word Problem Solver with Unified Tree Structure

A Neural Semantic Parser for Math Problems Incorporating Multi-Sentence Information

Solving Math Word Problems by Combining Language Models With Symbolic Solvers

A Knowledge-Aware Sequence-to-Tree Network for Math Word Problem Solving.

TDv2: A Novel Tree-Structured Decoder for Offline Mathematical Expression Recognition

Teacher-Student Networks with Multiple Decoders for Solving Math Word Problem

Enhancing Seq2seq Math Word Problem Solver with Entity Information and Math Knowledge