Neurosymbolic AI: the 3rd wave

Garcez, Artur d’Avila; Lamb, Luís C.

doi:10.1007/s10462-023-10448-w

Cited by 101 publications

(44 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…New directions that emphasize learning compositional models of real-world objects and events are being investigated in order to acquire more human levels of cognition [149]. Hybrid systems that borrow from the strengths of symbolic and deep learning paradigms are being actively pursued [155,161,162]. Leading AI experts have acknowledged the need for NLU systems to integrate knowledge at all levels of comprehension [149,163].…”

Section: Plos Onementioning

confidence: 99%

Design considerations for a hierarchical semantic compositional framework for medical natural language understanding

2023

View full text Add to dashboard Cite

Medical natural language processing (NLP) systems are a key enabling technology for transforming Big Data from clinical report repositories to information used to support disease models and validate intervention methods. However, current medical NLP systems fall considerably short when faced with the task of logically interpreting clinical text. In this paper, we describe a framework inspired by mechanisms of human cognition in an attempt to jump the NLP performance curve. The design centers on a hierarchical semantic compositional model (HSCM), which provides an internal substrate for guiding the interpretation process. The paper describes insights from four key cognitive aspects: semantic memory, semantic composition, semantic activation, and hierarchical predictive coding. We discuss the design of a generative semantic model and an associated semantic parser used to transform a free-text sentence into a logical representation of its meaning. The paper discusses supportive and antagonistic arguments for the key features of the architecture as a long-term foundational framework.

show abstract

Section: Plos Onementioning

confidence: 99%

Design considerations for a hierarchical semantic compositional framework for medical natural language understanding

2023

View full text Add to dashboard Cite

show abstract

“…On the symbolic side, we have explicit and reflective conceptual knowledge, often organized in causal models and which can be subject to reasoning. Since human intelligence is itself neuro-symbolic, there is no theoretical reason not to try to make the two branches of AI cooperate in hybrid artificial intelligence systems that combine learning and reasoning (Garcez and Lamb, 2020; D’Avila D’avila Garcez and Lamb, 2009; Lévy, 2021). The benefits are obvious and, in particular, each of the two subsystems can remedy the problems encountered by the other.…”

Section: Some Applicationsmentioning

confidence: 99%

Semantic computing with IEML

Lévy

2023

Collective Intelligence

View full text Add to dashboard Cite

This paper presents IEML, Information Economy MetaLanguage, a constructed language with the same expressive power as a natural language and with computable semantics. Distinguished from pragmatic and referential semantics, linguistic semantics have not yet been completely formalized. Only its syntagmatic dimension has been mathematized in the form of regular languages. Its paradigmatic dimension remained to be formalized. In order to complete the mathematizing of language, including its paradigmatic dimension, I have coded linguistic semantics with IEML. This article introduces its 3000-word dictionary, its formal grammar, and its integrated tools for building semantic graphs. For the future, IEML could become a vector for a fluid calculation and communication of meaning—semantic interoperability—capable of de-compartmentalizing the digital memory, and of advancing the progress of collective intelligence, artificial intelligence, and digital humanities. I conclude by indicating some research directions.

show abstract

“…Latent expressions can augment training signals with sub-symbolic features derived from explicit semantic content, but knowledge infusion per se doesn't determine how inference processes are conducted. Relevant work in this space shows how deep neural models can replicate logical reasoning (Ebrahimi, Eberhart, and Hitzler 2021; Garcez et al 2022), but it doesn't follow that any form of logical reasoning that is provably reducible to learning algorithms, should also be systematically reduced to it -this would be a requirement only for tightly-coupled neurosymbolic systems (Kautz 2022;Garcez and Lamb 2023). Accordingly, in the next section we make the case for developing an AI framework where the ACT-R architecture is loosely-coupled with neuro-symbolic components, to enable high-level reasoning.…”

Section: Motivationsmentioning

confidence: 99%

Chapter 13. Generalizable Neuro-Symbolic Systems for Commonsense Question Answering

Oltramari

Francis

Ilievski

et al. 2021

Frontiers in Artificial Intelligence and Applications

View full text Add to dashboard Cite

This chapter illustrates how suitable neuro-symbolic models for language understanding can enable domain generalizability and robustness in downstream tasks. Different methods for integrating neural language models and knowledge graphs are discussed. The situations in which this combination is most appropriate are characterized, including quantitative evaluation and qualitative error analysis on a variety of commonsense question answering benchmark datasets.

show abstract

Neurosymbolic AI: the 3rd wave

Cited by 101 publications

References 30 publications

Design considerations for a hierarchical semantic compositional framework for medical natural language understanding

Design considerations for a hierarchical semantic compositional framework for medical natural language understanding

Semantic computing with IEML

Chapter 13. Generalizable Neuro-Symbolic Systems for Commonsense Question Answering

Contact Info

Product

Resources

About