Multi-label, Multi-class Classification Using Polylingual Embeddings

Georgios Balikas (1, 2) , Massih-Reza Amini (1)

1 AMA - Analyse de données, Modélisation et Apprentissage automatique [Grenoble]
2 Coffreo

We propose a Polylingual text Embedding (PE) strategy, that learns a language independent representation of texts using Neu-ral Networks. We study the effects of bilingual representation learning for text classification and we empirically show that the learned representations achieve better classification performance compared to traditional bag-of-words and other monolingual distributed representations. The performance gains are more significant in the interesting case where only few labeled examples are available for training the classifiers.

Domaines: Intelligence artificielle [cs.AI]

Georgios Balikas, Massih-Reza Amini. Multi-label, Multi-class Classification Using Polylingual Embeddings. 38th European Conference on Information Retrieval ECIR 2016, Mar 2016, Padoue, Italy. ⟨10.1007/978-3-319-30671-1_59⟩. ⟨hal-01299850⟩