Cross-Media Sentiment Classification and Application to Box-Office Forecasting

Abstract : This article aims at demonstrating the interest of opinion mining on Twitter data for the box-office prediction. Whilst most approaches in box-office forecasting focus on expert knowledge (actor celebrity, film budget...), or more recently on Twitter volumetric features, we want to show that the tweet's content is also important to make an efficient decision. Firstly we focus on the cross-media sentiment classification task, by studying the impact different algorithms and data sources have on the accuracy of sentiment classification on Twitter. Secondly, models allow us to to build high level sentiment features for the box-office forecasting problem. We demonstrate the interest of opinion mining derived features for this second task.
Type de document :
Communication dans un congrès
the 10th Conference on Open Research Areas in Information Retrieval (OAIR '13), May 2013, Lisbon, Portugal. Proceedings of the 10th Conference on Open Research Areas in Information Retrieval (OAIR '13), pp.201-208, 2013
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01203008
Contributeur : Lip6 Publications <>
Soumis le : mardi 22 septembre 2015 - 10:35:18
Dernière modification le : dimanche 9 décembre 2018 - 01:22:05

Identifiants

  • HAL Id : hal-01203008, version 1

Collections

Citation

Elie Guardia-Sebaoun, Abdelhalim Rafrafi, Vincent Guigue, Patrick Gallinari. Cross-Media Sentiment Classification and Application to Box-Office Forecasting. the 10th Conference on Open Research Areas in Information Retrieval (OAIR '13), May 2013, Lisbon, Portugal. Proceedings of the 10th Conference on Open Research Areas in Information Retrieval (OAIR '13), pp.201-208, 2013. 〈hal-01203008〉

Partager

Métriques

Consultations de la notice

83