TU Darmstadt / ULB / TUbiblio

Neural, Multimodal, Energy-based Approach for Knowledge Graph Completion

Mousselly-Sergieh, Hatem ; Gurevych, Iryna ; Roth, Stefan (2017)
Neural, Multimodal, Energy-based Approach for Knowledge Graph Completion.
London, UK
Conference or Workshop Item, Bibliographie

Abstract

We propose an approach for knowledge graph (KG) completion that leverages multimodal information on KG entities including: 1) visual features which are obtained using state-of-the-art convolutional neural network models for image classification and 2) textual representations which are learned using word embedding techniques. Our approach builds upon the translation model (Bordes et al., 2013) for KG representation and introduce an energy-based framework that efficiently combines multimodal features. The corresponding KG representations are learned using a simple neural network architecture. We experimentally demonstrate the effectiveness of our approach and compare its performance to other baseline models.

Item Type: Conference or Workshop Item
Erschienen: 2017
Creators: Mousselly-Sergieh, Hatem ; Gurevych, Iryna ; Roth, Stefan
Type of entry: Bibliographie
Title: Neural, Multimodal, Energy-based Approach for Knowledge Graph Completion
Language: English
Date: September 2017
Book Title: Language-Learning-Logic Workshop (3L 2017)
Event Location: London, UK
Corresponding Links:
Abstract:

We propose an approach for knowledge graph (KG) completion that leverages multimodal information on KG entities including: 1) visual features which are obtained using state-of-the-art convolutional neural network models for image classification and 2) textual representations which are learned using word embedding techniques. Our approach builds upon the translation model (Bordes et al., 2013) for KG representation and introduce an energy-based framework that efficiently combines multimodal features. The corresponding KG representations are learned using a simple neural network architecture. We experimentally demonstrate the effectiveness of our approach and compare its performance to other baseline models.

Uncontrolled Keywords: CEDIFOR
Identification Number: TUD-CS-2017-0244
Divisions: 20 Department of Computer Science
20 Department of Computer Science > Ubiquitous Knowledge Processing
Date Deposited: 12 Sep 2017 16:36
Last Modified: 24 Jan 2020 12:03
PPN:
Corresponding Links:
Export:
Suche nach Titel in: TUfind oder in Google
Send an inquiry Send an inquiry

Options (only for editors)
Show editorial Details Show editorial Details