Abstract:Abstract This paper presents a relation-centric algorithm for solving arithmetic word problems (AWPs) by synergizing a syntax-semantics extractor for extracting explicit relations, and a neural network miner for mining implicit relations. This is the first algorithm that has a specific component to acquire implicit knowledge items for solving AWPs. This paper proposes a three-phase scheme to decompose the challenging task of designing an algorithm for solving AWPs into three smaller tasks. The first phase proposes a state-action paradigm; the second phase instantiates the paradigm into a relation-centric approach; and the third phase implements a relation-centric algorithm for solving AWPs. There are two main steps in the proposed algorithm: problem understanding and symbolic solver. By adopting the relation-centric approach, problem understanding becomes a task of relation acquisition. For conducting the task of relation acquisition, a relaxed syntax-semantics method first extracts a group of explicit relation candidates. In parallel, a neural network miner acquires implicit relation candidates. The miner computes the vectors encoded by BERT to determine which implicit relations should be added. Thus, problem understanding can acquire both explicit relations and implicit relations, which addresses the challenge of building a problem understanding method that can acquire all the knowledge items to find the solution. In the subsequent step of symbolic solver, a fusion procedure forms a distilled set of relations from all the candidates by discarding unnecessary relations. Experimentation on nine benchmark datasets validates the superiority of the proposed algorithm that outperforms the state-of-the-art algorithms.

Neural Solving Uninterpreted Predicates with Abstract Gradient Descent

Improving Neural Network Verification through Spurious Region Guided Refinement.

Learning Solving Procedure for Artificial Neural Network

Exploiting Problem Structure in Deep Declarative Networks: Two Case Studies

PDP: A General Neural Framework for Learning Constraint Satisfaction Solvers

MathDQN: Solving Arithmetic Word Problems Via Deep Reinforcement Learning.

Bridging Logic and Learning: A Neural-Symbolic Approach for Enhanced Reasoning in Neural Models (ASPER)

Deep Learning in Automatic Math Word Problem Solvers

Efficient compilation of expressive problem space specifications to neural network solvers

Learning to Solve Abstract Reasoning Problems with Neurosymbolic Program Synthesis and Task Generation

ASP: Learn a Universal Neural Solver!

Solving arithmetic word problems by synergizing syntax-semantics extractor for explicit relations and neural network miner for implicit relations

GLinSAT: The General Linear Satisfiability Neural Network Layer By Accelerated Gradient Descent

Progressive Interpretation Synthesis: Interpreting Task Solving by Quantifying Previously Used and Unused Information

Goal-Aware Neural SAT Solver

From Gameplay to Symbolic Reasoning: Learning SAT Solver Heuristics in the Style of Alpha(Go) Zero

DeepSaDe: Learning Neural Networks that Guarantee Domain Constraint Satisfaction

Techniques to Improve Neural Math Word Problem Solvers

Abstract Spatial-Temporal Reasoning Via Probabilistic Abduction and Execution

Interpretable Neural PDE Solvers using Symbolic Frameworks