Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Identification of Entailment and Contradiction Relations between Natural Language Sentences: A Neurosymbolic Approach

Xuyao Feng, Anthony Hunter

TL;DR

This work tackles the lack of explainability in natural language inference by proposing a neurosymbolic pipeline that translates sentences into Abstract Meaning Representation graphs, then into propositional logic, and finally uses SAT-based reasoning to determine entailment, contradiction, or neutrality. It introduces relaxation via neuro-matching and a forgetting mechanism to handle lexical variation and commonsense knowledge, enabling more robust entailment and contradiction detection. Evaluations on e-SNLI, SICK, and MultiNLI demonstrate that relaxation improves recall for entailment and contradiction and that explanations can boost performance on e-SNLI, though neutral detection remains challenging. The approach provides explicit, verifiable reasoning steps and a framework to incorporate commonsense via embedding-based relaxation, contributing to more transparent NLP inference systems.

Abstract

Natural language inference (NLI), also known as Recognizing Textual Entailment (RTE), is an important aspect of natural language understanding. Most research now uses machine learning and deep learning to perform this task on specific datasets, meaning their solution is not explainable nor explicit. To address the need for an explainable approach to RTE, we propose a novel pipeline that is based on translating text into an Abstract Meaning Representation (AMR) graph. For this we use a pre-trained AMR parser. We then translate the AMR graph into propositional logic and use a SAT solver for automated reasoning. In text, often commonsense suggests that an entailment (or contradiction) relationship holds between a premise and a claim, but because different wordings are used, this is not identified from their logical representations. To address this, we introduce relaxation methods to allow replacement or forgetting of some propositions. Our experimental results show this pipeline performs well on four RTE datasets.

Identification of Entailment and Contradiction Relations between Natural Language Sentences: A Neurosymbolic Approach

TL;DR

This work tackles the lack of explainability in natural language inference by proposing a neurosymbolic pipeline that translates sentences into Abstract Meaning Representation graphs, then into propositional logic, and finally uses SAT-based reasoning to determine entailment, contradiction, or neutrality. It introduces relaxation via neuro-matching and a forgetting mechanism to handle lexical variation and commonsense knowledge, enabling more robust entailment and contradiction detection. Evaluations on e-SNLI, SICK, and MultiNLI demonstrate that relaxation improves recall for entailment and contradiction and that explanations can boost performance on e-SNLI, though neutral detection remains challenging. The approach provides explicit, verifiable reasoning steps and a framework to incorporate commonsense via embedding-based relaxation, contributing to more transparent NLP inference systems.

Abstract

Natural language inference (NLI), also known as Recognizing Textual Entailment (RTE), is an important aspect of natural language understanding. Most research now uses machine learning and deep learning to perform this task on specific datasets, meaning their solution is not explainable nor explicit. To address the need for an explainable approach to RTE, we propose a novel pipeline that is based on translating text into an Abstract Meaning Representation (AMR) graph. For this we use a pre-trained AMR parser. We then translate the AMR graph into propositional logic and use a SAT solver for automated reasoning. In text, often commonsense suggests that an entailment (or contradiction) relationship holds between a premise and a claim, but because different wordings are used, this is not identified from their logical representations. To address this, we introduce relaxation methods to allow replacement or forgetting of some propositions. Our experimental results show this pipeline performs well on four RTE datasets.
Paper Structure (17 sections, 11 equations, 3 figures, 6 tables)

This paper contains 17 sections, 11 equations, 3 figures, 6 tables.

Table of Contents

  1. Literature review
  2. Background
  3. Abstract meaning representation
  4. Automated reasoning
  5. Datasets
  6. Method
  7. Entailment methods
  8. Propositional entailment
  9. Relaxed propositional entailment
  10. Contradiction methods
  11. Classification of relations
  12. Results
  13. Evaluation
  14. Different similarity thresholds
  15. Pipeline with forgetting versus without forgetting
  16. ...and 2 more sections

Figures (3)

  • Figure 1: AMR for the sentence "The boy wants to go." (left) and "The boy does not want to go." (right).
  • Figure 2: Our sentence-to-label pipeline. The text input is one or more premises and a claim, and the output is a label (entailment, contradiction, and neutral)
  • Figure 3: Results concerning the relaxation threshold. The y-axis represents the scores of precision, recall, F1-score, and accuracy, and the x-axis represents the threshold $\tau$.

Theorems & Definitions (26)

  • Example 1
  • Definition 1
  • Example 2
  • Definition 2
  • Definition 3
  • Example 3
  • Definition 4
  • Example 4
  • Definition 5
  • Example 5
  • ...and 16 more