Participation de l'IRISA à DeFT2017 : systèmes de classification de complexité croissante

Vincent Claveau 1 Christian Raymond 1
1 LinkMedia - Creating and exploiting explicit links between multimedia fragments
Inria Rennes – Bretagne Atlantique , IRISA_D6 - MEDIA ET INTERACTIONS
Abstract : This paper describes the participation the LinkMedia team from IRISA at DeFT2017. We carried out the three proposed tasks : classification of non-figurative tweets according to their polarity (task 1), identification figurative langage (task 2), classification of figurative and non-figurative tweets according to their polarity (task 3). For these three tasks, we adopt a standard supervised machine learning framework and explore the use of three methods of increasing complexity : i) k-nearest neighbors with information retrieval based techniques, ii) boosting of decision trees, iii) recurrent neural networks. Our approaches do not exploit rich external knowledge (lexicons, annotated corpora) and are chiefly based on the tweet content (and some other tweets for our latest approach). It allows us to evaluate the precise interest of each of our approach and the data representation that they use : bag-of-words for the first one, n-grams for the second and word embedding for the latest.
Liste complète des métadonnées

Cited literature [8 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01643993
Contributor : Vincent Claveau <>
Submitted on : Tuesday, November 21, 2017 - 7:06:15 PM
Last modification on : Monday, April 1, 2019 - 1:52:42 PM

File

DEFT2017.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01643993, version 1

Citation

Vincent Claveau, Christian Raymond. Participation de l'IRISA à DeFT2017 : systèmes de classification de complexité croissante. DeFT 2017 - Défi Fouille de texte, Jun 2017, Orléans, France. pp.1-10, 2017, Actes de l'Atelier Défi Fouille de Texte, DeFT. 〈hal-01643993〉

Share

Metrics

Record views

353

Files downloads

65