Vers un traitement automatique de la néosémie : approche textuelle et statistique

Abstract : The issue at stake is the automated meaning allocation. In a first time, a theoretical scheme is elaborated to describe meaning change for a lexical unit already defined in a lexical resource. We focus on semantic neology, considered as a significant repeted change. Our model relies on quantitative evidence and it is inspired from text semantics. The preexisting meaning is represented as a structured set of semantic features. The context modifies it due to salient semantic featuresin texts. These dynamic change is comprehended through description strata ranging from coarse-grained to fine-grained semantic units. In a second time, we dwell on relevant resources and tools from corpus linguistics. The resources are dictionaries and text corpus. Concretely, we use the Trésor de la Langue Française informatisé as a dictionary. Its entries are automatically converted into bags of semantic features. The textual data consists in three recent journalistic corpus. The resources are considered are mathematic spaces and statistical tools are used to extract significant units and to structure information. In a last time, we give an outline of a process to allocate automatically a new meaning. Experiments illustrate each step. This process relies on multiple levels of description, getting finer and finer. Through this approach, it is possible to qualify the new meaning in a precise and structured way.
Document type :
Theses
Complete list of metadatas

Cited literature [207 references]  Display  Hide  Download

https://hal.univ-lorraine.fr/tel-01749176
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 12:08:43 PM
Last modification on : Wednesday, April 11, 2018 - 1:28:06 AM
Long-term archiving on : Friday, September 14, 2018 - 10:16:49 AM

File

DDOC_T_2012_0038_REUTENAUER.pd...
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01749176, version 1

Collections

Citation

Coralie Reutenauer. Vers un traitement automatique de la néosémie : approche textuelle et statistique. Linguistique. Université de Lorraine, 2012. Français. ⟨NNT : 2012LORR0038⟩. ⟨tel-01749176⟩

Share

Metrics

Record views

37

Files downloads

85