Skip to Main content Skip to Navigation
Master Thesis

Exploitation de modèles neuronaux en vue de la détection automatique des nominations émergentes

Abstract : Emerging nominations are used to express the new concepts via new words or the words with semantic changes. They are taken for the precursor of neology insofar as they are not yet stabilized for use. Interested in the field of discourse analysis, they can convey implicitly the speaker's view and values, particularly about the social issues (e.g. integrated farming, sustainable mobility or selective immigration). By noticing the nominations in corpus, we observe that they are sometimes described by the contextual and formal marks, which could be exploited in detection. Base on this hypothesis, this work focuses on exploiting the deep learning models to automatically detect the criteria useful for the recognition of emerging nominations.
We propose a research base on a corpus which collect debates, press articles, and reports in the field of ecology for the period of 2018-2020. Firstly, we propose a study of detecting the specific contexts in order to pinpoint the emerging nominations. Secondly, we are interested in the FlauBERT’s MLM model for detecting the semantic changes via a difficulty predicting. In the end of work, the second criterion appears as the easiest to implement because our classifier produces too much noise.
Document type :
Master Thesis
Complete list of metadata
Contributor : Uga - Bulles Connect in order to contact the contributor
Submitted on : Friday, December 17, 2021 - 11:59:32 AM
Last modification on : Thursday, January 6, 2022 - 3:44:26 AM


DING Yumeng_M2_IDL.pdf
Files produced by the author(s)


  • HAL Id : dumas-03485460, version 1


Yumeng Ding. Exploitation de modèles neuronaux en vue de la détection automatique des nominations émergentes. Sciences de l'Homme et Société. 2021. ⟨dumas-03485460⟩



Les métriques sont temporairement indisponibles