Skip to Main content Skip to Navigation
Conference papers

Étude de l'influence des représentations textuelles sur la détection d'évènements non supervisée dans des flux de données

Abstract : Detection of real-world events using online data sources is a trending topic in the information retrieval domain. Multiple data sources are potentially of interest and some of them are data streams. There are multiple data sources that are potentially interesting, and some of them are textual data streams, structured or unstructured. We propose to analyse the problem of event detection from text data stream and to focus particularly on the importance of the representation of the textual data. To do so, we compare multiple approaches in different context: supervised and unsupervised.We focus on the performances of Transformer-based architectures for event detection on short text documents, and we conclude that, contrary to previous studies, these architectures can be competitive compared to classical methods.
Complete list of metadata

https://hal.archives-ouvertes.fr/hal-03324781
Contributor : Max Chevalier Connect in order to contact the contributor
Submitted on : Monday, August 30, 2021 - 9:51:27 AM
Last modification on : Tuesday, October 19, 2021 - 2:23:38 PM

Identifiers

  • HAL Id : hal-03324781, version 1

Citation

Elliot Maître, Zakaria Chemli, Max Chevalier, Bernard Douset, Jean-Philippe Gitto, et al.. Étude de l'influence des représentations textuelles sur la détection d'évènements non supervisée dans des flux de données. XXXIXème Congrès INFormatique des ORganisations et Systèmes d'Information et de Décision (INFORSID 2021), Jun 2021, Dijon (virtuel), France. pp.23-38. ⟨hal-03324781⟩

Share

Metrics

Record views

233

Files downloads

14