Do Convolutional Networks need to be Deep for Text Classification ?

Hoa Le 1 Christophe Cerisara 1 Alexandre Denis 2
1 SYNALP - Natural Language Processing : representations, inference and semantics
LORIA - NLPKD - Department of Natural Language Processing & Knowledge Discovery
Abstract : We study in this work the importance of depth in convolutional models for text classification, either when character or word inputs are considered. We show on 5 standard text classification and sentiment analysis tasks that deep models indeed give better performances than shallow networks when the text input is represented as a sequence of characters. However, a simple shallow-and-wide network outper-forms deep models such as DenseNet with word inputs. Our shallow word model further establishes new state-of-the-art performances on two datasets: Yelp Binary (95.9%) and Yelp Full (64.9%).
Document type :
Conference papers
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01690601
Contributor : Christophe Cerisara <>
Submitted on : Sunday, February 24, 2019 - 7:53:24 AM
Last modification on : Friday, March 1, 2019 - 3:54:57 PM

File

1707.04108.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01690601, version 1

Citation

Hoa Le, Christophe Cerisara, Alexandre Denis. Do Convolutional Networks need to be Deep for Text Classification ?. AAAI Workshop on Affective Content Analysis, Feb 2018, New Orleans, United States. ⟨hal-01690601⟩

Share

Metrics

Record views

240

Files downloads

22