Unsupervised Word Representations Learning with Bilinear Convolutional Network on Characters - Sorbonne Université
Conference Papers Year : 2021

Unsupervised Word Representations Learning with Bilinear Convolutional Network on Characters

Thomas Luka
  • Function : Author
  • PersonId : 1106646
Laure Soulier
David Picard

Abstract

In this paper, we propose a new unsupervised method for learning word embedding with raw characters as input representations, bypassing the problems arising from the use of a dictionary. To achieve this purpose, we translate the distributional hypothesis into a unsupervised metric learning objective, which allows to consider only an encoder instead of an encoder-decoder architecture. We propose to use a convolutional neural network with bilinear product blocks and residual connections to encode co-occurrences patterns. We show the efficiency of our approach by comparing it with classical word embedding methods such as fastText and GloVe on several benchmarks.

Dates and versions

hal-03479768 , version 1 (14-12-2021)

Identifiers

Cite

Thomas Luka, Laure Soulier, David Picard. Unsupervised Word Representations Learning with Bilinear Convolutional Network on Characters. The 29th European Symposium on Artificial Neural Networks, Oct 2021, Online, Belgium. pp.251-256, ⟨10.14428/esann/2021.ES2021-38⟩. ⟨hal-03479768⟩
48 View
0 Download

Altmetric

Share

More