Natural language syntax complies with the free-energy principle

Synthese, Journal Year: 2024, Volume and Issue: 203(5)

Published: May 3, 2024

Abstract Natural language syntax yields an unbounded array of hierarchically structured expressions. We claim that these are used in the service active inference accord with free-energy principle (FEP). While conceptual advances alongside modelling and simulation work have attempted to connect speech segmentation linguistic communication FEP, we extend this program underlying computations responsible for generating syntactic objects. argue recently proposed principles economy design—such as “minimal search” criteria from theoretical syntax—adhere FEP. This affords a greater degree explanatory power FEP—with respect higher functions—and offers linguistics grounding first computability. mostly focus on building new principled relations between also show through sample preliminary examples how both tree-geometric depth Kolmogorov complexity estimate (recruiting Lempel–Ziv compression algorithm) can be accurately predict legal operations workspaces, directly line formulations variational free energy minimization. is motivate general design term Turing–Chomsky Compression (TCC). use TCC align concerns linguists normative account self-organization furnished by marshalling evidence psycholinguistics ground core efficient computation within inference.

Language: Английский

---

DALL·E 2 fails to reliably capture common syntactic processes

Social Sciences & Humanities Open, Journal Year: 2023, Volume and Issue: 8(1), P. 100648 - 100648

Published: Jan. 1, 2023

Machine intelligence is increasingly being linked to claims about sentience, language processing, and an ability comprehend transform natural into a range of stimuli. We analyze the DALL·E 2 translate prompts images, since it remains unclear whether possesses facility for accurately representing grammatical strategies. target performance across 10 phenomena that are pervasive in human language: binding principles coreference, passives, structural ambiguity, negation, compositionality word order, quantification, double object constructions, sentence coordination, ellipsis, comparatives. In contrast young infants, who master these tasks, fails represent inferred meanings, performing at or near chance. While programs can be trained recognize vast numbers words calculate probabilities sequences, results challenge recent concerning artificial understanding language. The full set tested materials outputs made available as benchmark future testing.

Language: Английский

---

The spatiotemporal dynamics of semantic integration in the human brain

Nature Communications, Journal Year: 2023, Volume and Issue: 14(1)

Published: Oct. 24, 2023

Language depends critically on the integration of lexical information across multiple words to derive semantic concepts. Limitations spatiotemporal resolution have previously rendered it difficult isolate processes involved in integration. We utilized intracranial recordings epilepsy patients (n = 58) who read written word definitions. Descriptions were either referential or non-referential a common object. Semantically sentences enabled high frequency broadband gamma activation (70-150 Hz) inferior frontal sulcus (IFS), medial parietal cortex, orbitofrontal cortex (OFC) and temporal lobe left, language-dominant hemisphere. IFS, OFC posterior middle gyrus activity was modulated by coherence sentences, exposing effects that independent task-based status. Components this network, alongside superior sulcus, engaged for did not clearly reduce search space final word. These results indicate existence complementary cortical mosaics cortex.

Language: Английский

---

Speech and music recruit frequency-specific distributed and overlapping cortical networks

eLife, Journal Year: 2024, Volume and Issue: 13

Published: Feb. 8, 2024

To what extent does speech and music processing rely on domain-specific domain-general neural networks? Using whole-brain intracranial EEG recordings in 18 epilepsy patients listening to natural, continuous or music, we investigated the presence of frequency-specific network-level brain activity. We combined it with a statistical approach which clear operational distinction is made between shared, preferred, domain-selective responses. show that majority focal activity shared processing. Our data also reveal an absence anatomical regional selectivity. Instead, responses are restricted distributed coherent oscillations, typical spectral fingerprints. work highlights importance considering natural stimuli dynamics their full complexity map cognitive functions.

Language: Английский

---

Speech and music recruit frequency-specific distributed and overlapping cortical networks

eLife, Journal Year: 2024, Volume and Issue: 13

Published: July 22, 2024

To what extent does speech and music processing rely on domain-specific domain-general neural networks? Using whole-brain intracranial EEG recordings in 18 epilepsy patients listening to natural, continuous or music, we investigated the presence of frequency-specific network-level brain activity. We combined it with a statistical approach which clear operational distinction is made between shared , preferred, domain- selective responses. show that majority focal activity processing. Our data also reveal an absence anatomical regional selectivity. Instead, domain-selective responses are restricted distributed coherent oscillations, typical spectral fingerprints. work highlights importance considering natural stimuli dynamics their full complexity map cognitive functions.

Language: Английский

---

A comparative investigation of compositional syntax and semantics in DALL·E and young children

Social Sciences & Humanities Open, Journal Year: 2025, Volume and Issue: 11, P. 101332 - 101332

Published: Jan. 1, 2025

Language: Английский

---

Locating eloquent sites identified during brain tumor intraoperative mapping on reference MRI atlas

Communications Medicine, Journal Year: 2025, Volume and Issue: 5(1)

Published: May 7, 2025

Language: Английский

---

Anatomically distinct cortical tracking of music and speech by slow (1–8Hz) and fast (70–120Hz) oscillatory activity

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(5), P. e0320519 - e0320519

Published: May 8, 2025

Music and speech encode hierarchically organized structural complexity at the service of human expressiveness communication. Previous research has shown that populations neurons in auditory regions track envelope acoustic signals within range slow fast oscillatory activity. However, extent to which cortical tracking is influenced by interplay between stimulus type, frequency band, brain anatomy remains an open question. In this study, we reanalyzed intracranial recordings from thirty subjects implanted with electrocorticography (ECoG) grids left cerebral hemisphere, drawn existing open-access ECoG database. Participants passively watched a movie where visual scenes were accompanied either music or stimuli. Cross-correlation activity signals, along density-based clustering analyses linear mixed-effects modeling, revealed both anatomically overlapping functionally distinct mapping effect as function type band. We observed widespread left-hemisphere Slow Frequency Band (SFB, band-passed filtered low-frequency signal 1–8Hz), near zero temporal lags. contrast, High (HFB, 70–120Hz signal) was higher during perception, more densely concentrated classical language processing areas, showed frontal-to-temporal gradient lag values not perception musical Our results highlight complex interaction region band shapes dynamics naturalistic signals.

Language: Английский

---

Multiple dimensions of syntactic structure are resolved earliest in posterior temporal cortex

Progress in Neurobiology, Journal Year: 2024, Volume and Issue: 241, P. 102669 - 102669

Published: Sept. 25, 2024

Language: Английский

---

Speech and music recruit frequency-specific distributed and overlapping cortical networks

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2022, Volume and Issue: unknown

Published: Oct. 9, 2022

Abstract To what extent does speech and music processing rely on domain-specific domain-general neural networks? Using whole-brain intracranial EEG recordings in 18 epilepsy patients listening to natural, continuous or music, we investigated the presence of frequency-specific network-level brain activity. We combined it with a statistical approach which clear operational distinction is made between shared , preferred, domain- selective responses. show that majority focal activity processing. Our data also reveal an absence anatomical regional selectivity. Instead, domain-selective responses are restricted distributed coherent oscillations, typical spectral fingerprints. work highlights importance considering natural stimuli dynamics their full complexity map cognitive functions.

Language: Английский

---