文章背景与核心概要
本文介绍了一种用于论证分析的新型机器学习框架,该框架创新性地用“逻辑嵌入”(logical embeddings)取代了传统的上下文词嵌入。通过直接捕捉论证的逻辑语义与结构,该方法能够提供更优的语义表征,从而克服传统嵌入在处理复杂论证时的局限性。
为支撑这一框架,作者提出了一种基于数理逻辑的相似度度量方法,该方法具备严密的数学基础,能够满足标准余弦相似度所缺乏的关键理论性质。借助再生核希尔伯特空间(RKHS)理论,作者证明了这种最优编码方式可确保逻辑信息无损。实验表明,该方法在分类任务上的表现显著优于标准的嵌入技术。
Logical Embeddings for Argument Analysis
Summary
This paper introduces a novel machine-learning framework for argument analysis that replaces traditional contextualized word embeddings with logical embeddings. By directly capturing the logical semantics and argumentation structures of an argument, this approach offers superior meaning representation.
The framework is supported by a mathematically rigorous, logic-based similarity measure that guarantees theoretical properties lacking in standard cosine similarity-based embeddings. Leveraging the theory of Reproducing Kernel Hilbert Spaces (RKHS), the authors demonstrate that this optimal encoding prevents any loss of logical information. The proposed method has been implemented, benchmarked, and shown to outperform standard embedding techniques on classification tasks.
本文介绍了一种用于论证分析的新型机器学习框架,该框架用逻辑嵌入取代了传统的上下文词嵌入。通过直接捕捉论证的逻辑语义与论证结构,该方法提供了更优的语义表征。
该框架由一个数学严谨、基于逻辑的相似度度量提供支持,它保证了标准基于余弦相似度的嵌入所缺乏的理论属性。利用再生核希尔伯特空间(RKHS)的理论,作者证明了这种最优编码可防止任何逻辑信息的丢失。所提出的方法已经过实现和基准测试,并被证明在分类任务上优于标准的嵌入技术。
Document Metadata
| Field | Details |
|---|---|
| arXiv ID | arXiv:2608.15325 [cs.CL] |
| Primary Subject | Computation and Language (cs.CL) |
| Secondary Subjects | Artificial Intelligence (cs.AI) |
| Authors | Leander Heldring, Santiago Torres |
| Submission Date | August 15, 2026 |
| License | Creative Commons Attribution 4.0 International ![]() |
文档元数据
字段 详情 arXiv ID arXiv:2608.15325[cs.CL]主学科 计算与语言 ( cs.CL)次级学科 人工智能 ( cs.AI)作者 Leander Heldring, Santiago Torres 提交日期 2026年8月15日 许可协议 知识共享署名 4.0 国际版
Abstract
We propose a new framework for machine-learning-oriented argument analysis tasks. Our proposal involves replacing traditional contextualized word embeddings used in most NLP tasks with logical embeddings, an alternative encoding that directly exploits argumentation structures. In essence, logical embeddings encapsulate the logical semantics of an argument, allowing for a better representation of its meaning.
Supporting these embeddings is a mathematical logic-based similarity measure that offers a transparent notion of proximity and is guaranteed to satisfy several desirable theoretical properties that current cosine similarity-based contextualized word embeddings cannot assure. This similarity measure induces a positive semi-definite kernel on the set of arguments, enabling us to uniquely define logical embeddings using the theory of Reproducing Kernel Hilbert Spaces (RKHS).
Moreover, we prove that this encoding is optimal, in the sense that no logical information is lost in the process. As with other RKHS applications, logical embeddings can be used in numerous supervised and unsupervised tasks. We provide an implementation of the method and aim to test it against literature benchmarks. Additionally, we demonstrate that logical embeddings outperform most standard embedding methods on a classification task.
我们为面向机器学习的论证分析任务提出了一种新框架。我们的提议是用逻辑嵌入取代大多数自然语言处理(NLP)任务中使用的传统上下文词嵌入,这是一种直接利用论证结构的替代编码方式。本质上,逻辑嵌入概括了一个论证的逻辑语义,从而能更好地表征其含义。
支撑这些嵌入的是一种基于数理逻辑的相似度度量,它提供了一种透明的邻近度概念,并保证满足当前基于余弦相似度的上下文词嵌入所无法确保的几个理想理论属性。该相似度度量在论证集上导出一个半正定核,使我们能够利用再生核希尔伯特空间(RKHS)理论来唯一地定义逻辑嵌入。
此外,我们证明了这种编码是最优的,即在整个过程中没有任何逻辑信息丢失。与其他 RKHS 应用一样,逻辑嵌入可用于众多监督和无监督任务。我们提供了该方法的实现,并旨在对照文献基准对其进行测试。此外,我们证明了在分类任务中,逻辑嵌入的表现优于大多数标准的嵌入方法。
Access and Resources
- Full-Text Options:
- View PDF
- HTML Version (Experimental)
- TeX Source
- External Citations & Tools:
- Google Scholar
- Semantic Scholar
- NASA ADS
访问与资源
- 全文选项:
- 查看 PDF
- HTML 版本(实验性)
- TeX 源码
- 外部引用与工具:
- 谷歌学术
- Semantic Scholar
- NASA ADS
