Les modèles vectoriels de la mémoire sémantique : description, validation et perspectives

Abstract : "Latent Semantic Analysis" (LSA ; Landauer et Dumais, 1997) and "Hyperspace Analog to Language " (HAL ; Lund et Burgess, 1996) models can be called abstractive models (Tiberghien, 1997) because they model a semantic generalization over a number of learning episodes. LSA and HAL statistically analyze distribution of terms belonging to a large textual corpus to elaborate a semantic space in which each term is represented by a vector. The main goal of this note is to describe these two models by comparing them each other. We show that LSA and HAL are able to predict priming experiment data and to be combined with a comprehension model like Construction-Integration (Kintsch, 1988) to simulate in a realistic manner, signification access, predication process and construction of a consistent mental representation of a text.
Document type :
Journal articles
Complete list of metadatas

Cited literature [27 references]  Display  Hide  Download

https://hal.univ-cotedazur.fr/hal-01733925
Contributor : Pierre Thérouanne <>
Submitted on : Wednesday, March 14, 2018 - 4:15:49 PM
Last modification on : Monday, September 30, 2019 - 5:12:16 PM
Long-term archiving on : Wednesday, September 5, 2018 - 8:33:06 AM

File

CB PT GD 2004.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01733925, version 1

Collections

Citation

Cédrick Bellissens, Pierre Thérouanne, Guy Denhière. Les modèles vectoriels de la mémoire sémantique : description, validation et perspectives. Le Langage et l'Homme, EME éditions / L'Harmattan, 2004, 39 (101-122). ⟨hal-01733925⟩

Share

Metrics

Record views

116

Files downloads

167