Cognitive modelling with multilayer networks Insights advancements and future challenges Massimo Stella1 Salvatore Citraro23 Giulio

2025-04-27 0 0 4.99MB 48 页 10玖币

侵权投诉

Cognitive modelling with multilayer networks:

Insights, advancements and future challenges

Massimo Stella1*, Salvatore Citraro2,3, Giulio

Rossetti3, Daniele Marinazzo4, Yoed N. Kenett5and Michael

S. Vitevitch6

1*CogNosco Lab, Department of Computer Science, University of Exeter,

Exeter, UK.

2Department of Computer Science, University of Pisa, Pisa, Italy.

3Institute of Information Science and Technologies, National Research

Council, Pisa, Italy.

4Faculty of Psychology and Educational Sciences, Department of Data

Analysis, University of Ghent, Ghent, Belgium.

5Cognitive Complexity Lab, Faculty of Industrial Engineering and

Management, Technion, Israel Institute of Technology, Israel.

6Department of Psychology, University of Kansas, Lawrence, Kansas, USA.

*Corresponding author: massimo.stella@inbox.com;

Abstract

The mental lexicon is a complex cognitive system representing informa-

tion about the words/concepts that one knows. Decades of psychological

experiments have shown that conceptual associations across multiple,

interactive cognitive levels can greatly inﬂuence word acquisition, stor-

age, and processing. How can semantic, phonological, syntactic, and other

types of conceptual associations be mapped within a coherent mathemat-

ical framework to study how the mental lexicon works? We here review

cognitive multilayer networks as a promising quantitative and interpreta-

tive framework for investigating the mental lexicon. Cognitive multilayer

networks can map multiple types of information at once, thus capturing

how diﬀerent layers of associations might co-exist within the mental lex-

icon and inﬂuence cognitive processing. This review starts with a gentle

introduction to the structure and formalism of multilayer networks. We

then discuss quantitative mechanisms of psychological phenomena that

arXiv:2210.00500v1 [cs.CL] 2 Oct 2022

2Cognitive multilayer networks

could not be observed in single-layer networks and were only unveiled by

combining multiple layers of the lexicon: (i) multiplex viability highlights

language kernels and facilitative eﬀects of knowledge processing in healthy

and clinical populations; (ii) multilayer community detection enables con-

textual meaning reconstruction depending on psycholinguistic features;

(iii) layer analysis can mediate latent interactions of mediation, suppres-

sion and facilitation for lexical access. By outlining novel quantitative per-

spectives where multilayer networks can shed light on cognitive knowledge

representations, also in next-generation brain/mind models, we discuss

key limitations and promising directions for cutting-edge future research.

Keywords: Cognitive modelling, multilayer networks, multiplex networks,

cognition, knowledge modelling, cognitive data science.

1 Introduction

The mental lexicon is a complex system where knowledge of the words and

concepts one knows can be represented as units that are combined and associ-

ated across multiple levels [1]. For example, phonemes combine to form words,

words combined in sentences express ideas, and sentences in narratives give rise

to stories [2,3]. Focusing on the level of units of words (which provide meaning

even in isolation), deeper knowledge can be expressed by linking together units

that are associated in some way. Words can be associated in many ways [1,4,5].

For example, words may share meaning [6], sound similar [7], be syntactically

related [8], bring each other to mind [9], represent objects with similar semantic

or visual features [10], be written similarly [11] or evoke the same set of emo-

tions and aﬀective states [12]. These are only some of the many ways in which

words can be associated [1,2,13] and give structure to the knowledge that one

has that can be expressed through language. Decades of research in psycholin-

guistics and cognitive science have examined how the words and concepts in the

mental lexicon are acquired, stored, processed, and retrieved [1,5,14]. Impor-

tantly, it has been shown that the structure and organisation of the words and

Cognitive multilayer networks 3

concepts associated in some way in the mental lexicon inﬂuence a wide variety

of linguistic and cognitive phenomena, such as word confusability [2], picture

naming [15–17], and memory recall patterns for both neutral [9,18–20] and

emotional information [21,22]. The structure and organisation of the words and

concepts associated in some way in the mental lexicon can be inﬂuenced by var-

ious factors, including psychedelic drugs [23], and how creative [24], expert [25]

or curious [26] an individual is. All these ﬁndings converge on one point: Under-

standing the structure and organisation of knowledge in the mental lexicon is

important for shedding light on a number of phenomena. Understanding the

structure and organisation of knowledge in the mental lexicon requires a frame-

work that is quantitative [27], interpretable [28] and human-centric [29]. This

framework must: (i) be capable of producing inferences and comparable mea-

surements regulated by mathematical equations and theoretical models [16,30]

(quantitative); (ii) map results to outputs through an internal representation

of knowledge available to researchers, unlike most black-box machine learning

knowledge models [21] (interpretable); and (iii) be grounded in psychological

theory and large-scale datasets in order to account for the complex nuances of

human psychology rather than make abstract inferences that are of little value

to psychologists [31,32]. An artiﬁcial intelligence that categorises individuals

using binary labels like “aphasic” or “healthy” without identifying the sever-

ity of their language impairments, nor considers their ability to acquire, retain,

and produce new knowledge would not be human-centric [17]).

In the present review, we advance the idea of using multilayer networks to

model and understand the structure and organisation of knowledge in the men-

tal lexicon. We discuss recent work from multiple ﬁelds to show how multilayer

networks are a quantitative, interpretable, and human-centric framework that

can connect several disparate disciplines interested in modelling knowledge.

4Cognitive multilayer networks

Multilayer networks are a cutting-edge approach to explore how knowledge is

processed simultaneously across multiple levels. We outline 3 recent research

developments where the ability to combine diﬀerent layers of associative knowl-

edge highlights phenomena that would be otherwise lost in single-layer network

analyses or through other modelling approaches like word embeddings [28].

We discuss key limitations of this framework and review potential approaches

for future research in cognitive modelling [33,34] and cognitive neuroscience

[31,35]. Combining evidence from ﬁelds as diverse yet interconnected as

cognitive psychology, complexity science, and computer science, our review

identiﬁes concrete innovative ways in which multilayer networks can advance

our understanding of cognition.

2 Evidence for the multilayered nature of the

mental lexicon

Despite the name, the mental lexicon is not a simple dictionary [2,5,36,37].

Concepts in the mental lexicon are not recalled in alphabetical order and the

recollection of an item is not independent of other concepts associated with

it [38,39]. Aitchison [1] used the London tube as a metaphor for the mental

lexicon, where stations represent linguistic units and are connected according

to a layout of channels of diﬀerent lengths. This analogy resonates with the

concept of a complex network, although the exact speciﬁcation of structure,

function, and dynamics in the mental lexicon is more sophisticated [37]. Even

though the mental lexicon might not be a network itself, some of its associative

features might be accurately modelled by network science [39].

Many research ﬁndings indicate that information represented in the men-

tal lexicon is inherently multi-layered: Phonological, semantic, and syntactic

Cognitive multilayer networks 5

aspects of language can simultaneously interact and inﬂuence language retrieval

and processing [5,16,40]. In healthy populations, the interaction of multi-

ple types of linguistic interactions in the mental lexicon is highlighted by the

phenomenon of the tip of the tongue [5], where an individual is aware of

the semantic features of a word but cannot produce it. This tip-of-the-tongue

state is characterised by a failure to retrieve phonological information, whereas

semantic activation seems to be intact [36,41]. Another example of faulty

retrieval is known as a malapropism [36,42] where a similar sounding word is

retrieved for another semantically appropriate one (e.g. “dancing a ﬂamingo”

instead of “dancing a ﬂamenco”). The faulty interaction between semantic and

phonological information of words can also explain the increase of mixed errors

in people with aphasia in a picture naming task [40].

Evidence for the multilayered nature of the mental lexicon comes also from

facilitative eﬀects in word production like priming [1,19,38], i.e. when lexical

retrieval is facilitated by cues related to target words. Morphological content

(e.g. “dog” containing phonemes \d\,\o\and \g\), synonym similarities (e.g.

“character” and “nature” being synonyms), and syntactic relationships (e.g.

being a certain part of speech) were found to facilitate lexical retrieval through

priming indicating the simultaneous interplay between phonological, semantic,

and syntactic layers of the mental lexicon [37,43]. These ﬁndings motivated

the formulation of the so-called cognitive linguistic theory [44], of serial lexical

access [40] and of cobweb theory [1], which all argue that language produc-

tion depends on a network of interacting layers of the mental lexicon, including

individual phonemes, word meaning, and sentence structuring. Given the inter-

action of various types of information in the lexicon, the framework of multilayer

networks becomes a natural way of analysing the structural and dynamical

complexity of the mental lexicon.

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

Cognitivemodellingwithmultilayernetworks:Insights,advancementsandfuturechallengesMassimoStella1*,SalvatoreCitraro2,3,GiulioRossetti3,DanieleMarinazzo4,YoedN.Kenett5andMichaelS.Vitevitch61*CogNoscoLab,DepartmentofComputerScience,UniversityofExeter,Exeter,UK.2DepartmentofComputerScience,UniversityofPi...

展开>> 收起<<

Cognitive modelling with multilayer networks Insights advancements and future challenges Massimo Stella1 Salvatore Citraro23 Giulio.pdf

共48页,预览5页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Cognitive modelling with multilayer networks Insights advancements and future challenges Massimo Stella1 Salvatore Citraro23 Giulio

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: