• Aucun résultat trouvé

Dictionary, Grounding Kernel, Kernel Core and Hierarchy

N/A
N/A
Protected

Academic year: 2021

Partager "Dictionary, Grounding Kernel, Kernel Core and Hierarchy"

Copied!
2
0
0

Texte intégral

(1)

Dictionary, Grounding Kernel, Kernel Core and Hierarchy

GK KC apple bad banana color dark edible fruit good light no not red tomato yellow apple 4 bad 1 banana 4 color 2 dark 1 edible 2 fruit3 good 1 light 1 no 0 not 0 red 3 tomato 4 yellow 3

left: Compressing an artificially constructed mini-dictionary:

First remove all words reachable by definition alone. That leaves Grounding Kernel (GK) and Kernel Core (KC).

right: Definitional distance hierarchy:

KC is 0level. Furthest definitional distance is4

(2)

Results

hi hi hilohi lo lo lo hi lo Concreteness Imageability Spoken Frequency Age of acquisition Written Frequency GK KC D 0 1 2 3 4 5 6 7 8 100 200 300 400 500 Level Value AOA SF C I WF

left: Summary schema for compressed real dictionary:

Definitional distance is correlated with 5 psycholinguistic variables. Words in KC are acquired at younger age.

Correlation with age and oral/written frequency is dichotomous.

Correlation with concreteness and imageability is continuous across all 8 levels of the definitional distance hierarchy.

right: Mean values for 5 psycholinguistic correlates

The 3 dichotomous correlates (AOA, WF, SF) and 2 continuous correlates (C, I) at each definitional distance. (8th level spurious; sample too small.) O. Picard et al. (ISC, LaCIM, GERAD) Hierarchies in Dictionary Definition Space NIPS-Graphs 2009 2 / 2

Références

Documents relatifs

Alternatively, Alaoui and Mahoney [1] showed that sampling columns according to their ridge leverage scores (RLS) (i.e., a measure of the influence of a point on the

Combined with Sch¨ utzen- berger’s characterization of star-free languages in terms of finite aperiodic monoids [22], this leads to an algorithm to decide whether a given

In this paper we investigate online kernel adaptive algorithms to identify Multiple Inputs Multiple Outputs model as well as the possibility of dictionary adaptation for such

This version of gCCA captures linear relationships between blocks and to assess nonlinear relations, we propose in this paper a kernel extension of gCCA.. The paper is organized

The problem of inexact graph matching is twofold : first find a similarity measure between graphs of different sizes, then find the best match between graphs in an “acceptable”

We present basic structures, definitions, normal forms, and a hierarchy of languages based on catenation, shuffle and their itera- tions, defined by algebraic closure or least fix

This paper evaluates the mitigation of STT-MRAM performance obstacle by ex- ploiting its density to increase LLC cache size and by applying an enhanced cache replacement policy