FrenchSentiClass : un Système Automatisé pour la Classification de Sentiments en Français

Abstract : This paper describes the system we used on the tasks of the text mining challenge (DEFT 2017). This thirteenth edition of this challenge concerned the analysis of opinions and figurative language in French tweets. Three tasks have been proposed : (i) the first one concerns the classification of non-figurative tweets according to their polarity ; (ii) the second one concerns the identification of figurative language, while (iii) the third one concerns the classification of figurative and non-figurative tweets according to their polarity. We proposed an automated system based on Support Vector Machines (SVM). The system automatically chooses on each step the best preprocessing, syntactic features and sentiment lexicons by cross validation on the training set. Furthermore, it performs an evaluation of feature subset selection and a tuning SVM complexity parameter. Therefore, this system can significantly reduce the time necessary to explore the data and choose the best feature representation.
Complete list of metadatas

Cited literature [15 references]  Display  Hide  Download

https://hal-lirmm.ccsd.cnrs.fr/lirmm-01563411
Contributor : Amine Abdaoui <>
Submitted on : Monday, July 17, 2017 - 4:59:33 PM
Last modification on : Wednesday, July 3, 2019 - 3:24:08 PM
Long-term archiving on : Saturday, January 27, 2018 - 3:04:12 AM

File

deft2017-frenchsenticlassEquip...
Files produced by the author(s)

Identifiers

  • HAL Id : lirmm-01563411, version 1

Citation

Mike Donald Tapi Nzali, Amine Abdaoui, Jérôme Azé, Sandra Bringay, Christian Lavergne, et al.. FrenchSentiClass : un Système Automatisé pour la Classification de Sentiments en Français. DEFT: Défi Fouille de Texte, Jun 2017, Orléans, France. ⟨lirmm-01563411⟩

Share

Metrics

Record views

495

Files downloads

434