Skip to Main content Skip to Navigation

Le rôle des inférences pour la fouille d'opinion : applications aux réseaux sociaux en langue chinoise

Abstract : This thesis is interested in linguistic inference in opinion mining in a corpus of tourist commentaries in Chinese. Existing techniques which are well developed on short and explicit opinions, give limited results in interpreting implicit contexts. In addition, expression of opinion implements different enunciative strategies according to languages ​​and cultures. Our hypothesis consists in studying inferences to improve opinion mining. In this perspective, our first contribution proposes a typology of inferences for Chinese in 5 types: logical, pragmatic, lexical, enunciative and discursive (Rossi and Campion, 1999; Marin, 2004; Duchêne, 2008; Doucy and Massoussi, 2012). We applied this typology to annotate a corpus, with the objective of conducting opinion mining experiments with and without the processing of inferences. Our second contribution focuses on automatic classification of inferences based on linguistic characteristics, domain metadata and word embedding vectors. The objective on the one hand is to prove that the processing of inferences improves the performance of opinion mining and on the other hand to find a balanced solution between expensive manual annotation and automatic classification. In this thesis, we demonstrated the interest of studying inferences for opinion mining in Chinese. However, the automatic identification of inferences remains complex and requires further research.
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Tuesday, December 7, 2021 - 5:53:08 PM
Last modification on : Friday, January 21, 2022 - 3:31:29 AM


Version validated by the jury (STAR)


  • HAL Id : tel-03469568, version 1



Liyun Yan. Le rôle des inférences pour la fouille d'opinion : applications aux réseaux sociaux en langue chinoise. Linguistique. Institut National des Langues et Civilisations Orientales- INALCO PARIS - LANGUES O', 2021. Français. ⟨NNT : 2021INAL0016⟩. ⟨tel-03469568⟩



Record views


Files downloads