A mechanism for the cortical computation of hierarchical linguistic structure.
A mechanism for the cortical computation of hierarchical linguistic structure.
复制标题
DOI：
10.1371/journal.pbio.2000663
复制
发表时间：
2017-03
期刊：
影响因子：
9.8
通讯作者：
Doumas LA
中科院分区：
文献类型：
--
作者：
Martin AE;Doumas LA
生物系统经常检测环境中的物种特异性信号。在人类中，言语和语言是具有基本生物重要性的物种特异性信号。为了检测语言信号，人类大脑必须从一系列随时间分布的感知输入中形成分层表示。这种能力背后的机制是什么？一种假设是，当层次语言表征成为有机体面临的计算问题的有效解决方案时，大脑重新利用了一种可用的神经生物学机制。在这种情况下，单一机制必须具有执行多个功能相关计算的能力，例如，检测语言信号并执行其他认知功能，同时，理想情况下，像人脑一样振荡。我们表明，一个计算模型的类比，建立了一个完全不同的目的学习关系推理过程的句子，代表他们的意义，而且，至关重要的是，表现出类似于由相同的刺激引起的皮层信号的振荡激活模式。皮质和机器信号中的这种冗余表明代表性结构构建和“皮质”振荡之间的形式和机械对准。通过归纳推理，这种协同作用表明皮层信号反映了结构的产生，就像机器信号一样。一个单一的机制-使用时间来编码信息在分层的网络-产生的（去）组成的代表性层次结构，这是人类语言的关键，并提供了一个机械的语言表征和皮层计算之间的联系假设。人类语言是一种基本的生物信号，它具有不同于其他感知-动作系统的计算特性：声音、单词、短语和句子之间的层次关系，以及将较小的单元联合收割机组合成较大单元的无限能力，从而产生“离散无限”的表达。长期以来，这些特性使得语言很难从生物系统的角度和认知模型中解释。我们认为，一个单一的计算机制，使用时间来编码层次，可以满足计算的语言要求，除了那些其他的认知功能。我们证明了一个支持良好的类比神经网络模型在处理句子时会像人脑一样振荡。尽管是为了完全不同的目的（学习关系概念）而构建的，但该模型处理句子的分层表示，并表现出与人类皮层对相同刺激的反应非常相似的振荡激活模式。从该模型中，我们得出了一个明确的计算机制，人类大脑如何将感知特征转换为跨多个时间尺度的层次表示，提供了语言和皮层计算之间的联系假设。我们的研究结果表明，一个正式的和机械的代表性的结构建设和皮层振荡之间的对齐，具有广泛的影响，发现在人类大脑中的认知计算的第一原则。
Biological systems often detect species-specific signals in the environment. In humans, speech and language are species-specific signals of fundamental biological importance. To detect the linguistic signal, human brains must form hierarchical representations from a sequence of perceptual inputs distributed in time. What mechanism underlies this ability? One hypothesis is that the brain repurposed an available neurobiological mechanism when hierarchical linguistic representation became an efficient solution to a computational problem posed to the organism. Under such an account, a single mechanism must have the capacity to perform multiple, functionally related computations, e.g., detect the linguistic signal and perform other cognitive functions, while, ideally, oscillating like the human brain. We show that a computational model of analogy, built for an entirely different purpose—learning relational reasoning—processes sentences, represents their meaning, and, crucially, exhibits oscillatory activation patterns resembling cortical signals elicited by the same stimuli. Such redundancy in the cortical and machine signals is indicative of formal and mechanistic alignment between representational structure building and “cortical” oscillations. By inductive inference, this synergy suggests that the cortical signal reflects structure generation, just as the machine signal does. A single mechanism—using time to encode information across a layered network—generates the kind of (de)compositional representational hierarchy that is crucial for human language and offers a mechanistic linking hypothesis between linguistic representation and cortical computation. Human language is a fundamental biological signal with computational properties that differ from other perception-action systems: hierarchical relationships between sounds, words, phrases, and sentences and the unbounded ability to combine smaller units into larger ones, resulting in a "discrete infinity" of expressions. These properties have long made language hard to account for from a biological systems perspective and within models of cognition. We argue that a single computational mechanism—using time to encode hierarchy—can satisfy the computational requirements of language, in addition to those of other cognitive functions. We show that a well-supported neural network model of analogy oscillates like the human brain while processing sentences. Despite being built for an entirely different purpose (learning relational concepts), the model processes hierarchical representations of sentences and exhibits oscillatory patterns of activation that closely resemble the human cortical response to the same stimuli. From the model, we derive an explicit computational mechanism for how the human brain could convert perceptual features into hierarchical representations across multiple timescales, providing a linking hypothesis between linguistic and cortical computation. Our results suggest a formal and mechanistic alignment between representational structure building and cortical oscillations that has broad implications for discovering the computational first principles of cognition in the human brain.