Analyse automatique d'opinion : problématique de l'intensité et de la négation pour l'application à un corpus journalistique

Lei Zhang 1
1 Equipe Hultech - Laboratoire GREYC - UMR6072
GREYC - Groupe de Recherche en Informatique, Image, Automatique et Instrumentation de Caen
Abstract : The field of opinion analysis has a rising interest since fifteen years. This field gathers numerous studies from the constitution of lexicon to opinion holder and target spotting. The objective of our study is automatic opinion analysis in news for economic intelligence. This objective is ambitious beside the state of the art as we must study many criteria concerning opinion analysis. We propose a conceptual model allowing the characterization of opinion according to polarity, intensity, prototypicality, target, source, author engagement, etc. We have specifically designed a model for the negation and its effects on intensity and polarity. In addition to the classical effect of negation, our model is able to manage rhetorical usages such as euphemism which is frequently used in news texts. Based on this conceptual model, we set technical aspects allowing to define the data and processing required for implementation. We developed two applications (in Chinese and in French) and conclude in the validity of our conceptual model as well as the sustainability of the computation.
Document type :
Theses
Complete list of metadatas

Cited literature [61 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00777603
Contributor : Lei Zhang <>
Submitted on : Thursday, January 17, 2013 - 4:54:44 PM
Last modification on : Tuesday, February 5, 2019 - 12:12:40 PM
Long-term archiving on: Saturday, April 1, 2017 - 6:50:05 AM

Identifiers

  • HAL Id : tel-00777603, version 1

Citation

Lei Zhang. Analyse automatique d'opinion : problématique de l'intensité et de la négation pour l'application à un corpus journalistique. Traitement du texte et du document. Université de Caen, 2012. Français. ⟨tel-00777603⟩

Share

Metrics

Record views

1311

Files downloads

3673