Lune

EMNLP Findings2021

Neural Unification for Logic Reasoning over Natural Language

Gabriele Picco, Thanh Lam Hoang, Marco Luca Sbodio, Vanessa López

Abstract

Automated Theorem Proving (ATP) deals with the development of computer programs being able to show that some conjectures (queries) are a logical consequence of a set of axioms (facts and rules). There exists several successful ATPs where conjectures and axioms are formally provided (e.g. formalised as First Order Logic formulas). Recent approaches, such as (Clark et al., 2020), have proposed transformer-based architectures for deriving conjectures given axioms expressed in natural language (English). The conjecture is verified through a binary text classifier, where the transformers model is trained to predict the truth value of a conjecture given the axioms. The RuleTaker approach of (Clark et al., 2020) achieves appealing results both in terms of accuracy and in the ability to generalize, showing that when the model is trained with deep enough queries (at least 3 inference steps), the transformers are able to correctly answer the majority of queries (97.6%) that require up to 5 inference steps. In this work we propose a new architecture, namely the Neural Unifier, and a relative training procedure, which achieves state-of-the-art results in term of generalisation, showing that mimicking a well-known inference procedure, the backward chaining, it is possible to answer deep queries even when the model is trained only on shallow ones. The approach is demonstrated in experiments using a diverse set of benchmark data. The source code is available at this location 1 .

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

2 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • ProofWriter: Generating Implications, Proofs, and Abductive Statements over Natural Language cites PRover: Proof Generation for Interpretable Reasoning over Rules
  • Explicit Planning Helps Language Models in Logical Reasoning cites ProofWriter: Generating Implications, Proofs, and Abductive Statements over Natural Language
  • Explicit Planning Helps Language Models in Logical Reasoning cites PRover: Proof Generation for Interpretable Reasoning over Rules
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers2

  1. Learning Deductive Reasoning from Synthetic Corpus based on Formal LogicTerufumi Morishita, Gaku Morio, Atsuki Yamaguchi, Yasuhiro SogawaICML 2023 · 45 citations
  2. Explicit Planning Helps Language Models in Logical ReasoningHongyu Zhao, Kangrui Wang, Mo Yu, Hongyuan MeiEMNLP 2023 · 8 citations

Builds on4

  1. ProofWriter: Generating Implications, Proofs, and Abductive Statements over Natural LanguageOyvind Tafjord, Bhavana Dalvi, Peter ClarkACL Findings 2021 · 122 citations
  2. Differentiable Reasoning on Large Knowledge Bases and Natural LanguagePasquale Minervini, Matko Bosnjak, Tim Rocktäschel, Sebastian Riedel et al.AAAI 2020 · 94 citations
  3. Differentiable Reasoning over a Virtual Knowledge BaseBhuwan Dhingra, Manzil Zaheer, Vidhisha Balachandran, Graham Neubig et al.ICLR 2020 · 91 citations
  4. PRover: Proof Generation for Interpretable Reasoning over RulesSwarnadeep Saha, Sayan Ghosh, Shashank Srivastava, Mohit BansalEMNLP 2020 · 3 citations

Related papers

  1. Proving Theorems using Incremental Learning and Hindsight Experience ReplayEser Aygün, Ankit Anand, Laurent Orseau, Xavier Glorot et al.ICML 2022 · 22 citations
  2. HyperTree Proof Search for Neural Theorem ProvingGuillaume Lample, Timothée Lacroix, Marie-Anne Lachaux, Aurélien Rodriguez et al.NeurIPS 2022 · 271 citations
  3. Measuring Systematic Generalization in Neural Proof Generation with TransformersNicolas Gontier, Koustuv Sinha, Siva Reddy, Christopher PalNeurIPS 2020 · 69 citations
  4. Learning to Prove Theorems by Learning to Generate TheoremsMingzhe Wang, Jia DengNeurIPS 2020 · 60 citations
  5. BAIT: Benchmarking (Embedding) Architectures for Interactive Theorem-ProvingSean Lamont, Michael Norrish, Amir Dezfouli, Christian Walder et al.AAAI 2024 · 3 citations
Dusk over the sea between two cliffs drawn in fine vertical lines