Implementing a concept network model

Abstract

The same concept can mean different things or be instantiated in different forms depending on context, suggesting a degree of flexibility within the conceptual system. We propose that a compositional network model can be used to capture and predict this flexibility. We modeled individual concepts (e.g., BANANA, BOTTLE) as graph-theoretical networks, in which properties (e.g., YELLOW, SWEET) were represented as nodes and their associations as edges. In this framework, networks capture the within-concept statistics that reflect how properties correlate with each other across instances of a concept. We ran a classification analysis using graph eigendecomposition to validate these models, and find that these models can successfully discriminate between object concepts. We then computed formal measures from these concept networks and explored their relationship to conceptual structure. We find that diversity coefficients and core-periphery structure can be interpreted as network-based measures of conceptual flexibility and stability, respectively. These results support the feasibility of a concept network framework and highlight its ability to formally capture important characteristics of the conceptual system. BANANA: banana chips, banana pudding, Cavendish banana, fried banana, frozen banana, mashed banana, over-ripe banana, peeled banana, plantain, raw banana, red banana, rotten banana, sliced banana, unripe banana, ripe banana BOTTLE: baby bottle, beer bottle, broken bottle, juice bottle, liquor bottle, medicine bottle, milk bottle, soda bottle, spray bottle, water bottle, wine bottle KEY: car key, key to a city, door key, encryption key, garage key, key to my heart, house key, key card, keyboard key, map key, master key, motorcycle key, office key, padlock key, password, piano key, key to a safe, skeleton key PUMPKIN: pumpkin bar, pumpkin bread, pumpkin candle, canned pumpkin, pumpkin cookie, pumpkin in a field, Halloween pumpkin, Jack-O-Lantern, pumpkin latte, pumpkin muffin, pumpkin pie, pumpkin puree, rotten pumpkin, pumpkin seeds, smashed pumpkin, pumpkin soup, pumpkin spice, whole pumpkin, Thanksgiving pumpkin

Cognitive Network Science: A Review of Research on Cognition through the Lens of Network Representations, Processes, and Dynamics

“…Associative strength is one of many options to construct networks of words. Edges between words can also be constructed based on the number of shared features (e.g., the concepts banana and cheese are connected as they are both yellow in color; [42], see also [43]), their semantic relations, such as synonymy (e.g., happy and joy share similar meanings), hypernymy (e.g., maple is a tree), and meronymy (e.g., a bird has a beak; see [44]), their phonological similarity (i.e., words that sound similar are connected to each other; [32,45,46]), their orthographic similarity (i.e., words that have similar spellings are connected to each other; [47,48]), their cooccurrences in naturalistic speech [49], language corpora statistics [50], and manually annotated syntactic dependency relationships ( [51,52]; see [53], for a review). Likewise, research has studied different linguistic units other than words by mapping letters [54,55], syllables or segments [56], or entire documents [57] onto nodes.…”

High-level cognition during story listening is reflected in high-order dynamic correlations in neural activity patterns

“…For example, prior work has used region of interest analyses to estimate the anatomical locations of specific neural representations 10 , or to compare the relative contributions to the neural code of multivariate activity patterns versus dynamic correlations between neural activity patterns 11,12 . An emerging theme in this literature is that cognition is mediated by dynamic interactions between brain structures [13][14][15][16][17][18][19][20][21][22][23][24][25] .…”

High-level cognition during story listening is reflected in high-order dynamic correlations in neural activity patterns

“…We tested the hypothesis that high-level cognition is reflected in high-order brain network dynamics (e.g., see Reimann et al, 2017; Solomon et al, 2019). We examined high-order network dynamics in functional neuroimaging data collected during a story listening experiment.…”

“…An emerging theme in this literature is that cognition is mediated by dynamic interactions between brain structures (Bassett et al, 2006;Bressler & Kelso, 2001;Demertzi et al, 2019;Friston, 2000;Grossberg, 1988;Lurie et al, 2018;Mack et al, 2017;McIntosh, 2000;Preti et al, 2017;Solomon et al, 2019;Sporns & Honey, 2006;Turk-Browne, 2013;Zou et al, 2019).…”

“…For example, prior work has used region of interest analyses to estimate the anatomical locations of specific neural representations (e.g., Etzel et al, 2009), or to compare the relative contributions to the neural code of multivariate activity patterns versus dynamic correlations between neural activity patterns (e.g., Fong et al, 2019; Manning et al, 2018). An emerging theme in this literature is that cognition is mediated by dynamic interactions between brain structures (Bassett et al, 2006; Demertzi et al, 2019; Friston, 2000; Grossberg, 1988; Lurie et al, 2018; Mack et al, 2017; Preti et al, 2017; Solomon et al, 2019; Sporns & Honey, 2006; Turk-Browne, 2013; Zou et al, 2019). Studies of the neural code to date have primarily focused on univariate or multivariate neural patterns (for review see Norman et al, 2006), or (more recently) on patterns of dynamic first-order correlations (i.e., interactions between pairs of brain structures; Demertzi et al, 2019; Fong et al, 2019; Lurie et al, 2018; Manning et al, 2018; Preti et al, 2017; Zou et al, 2019).…”