Alert button

Chains of Reasoning over Entities, Relations, and Text using Recurrent Neural Networks

Add code
Bookmark button
Alert button
May 01, 2017
Rajarshi Das, Arvind Neelakantan, David Belanger, Andrew McCallum

Figure 1 for Chains of Reasoning over Entities, Relations, and Text using Recurrent Neural Networks
Figure 2 for Chains of Reasoning over Entities, Relations, and Text using Recurrent Neural Networks
Figure 3 for Chains of Reasoning over Entities, Relations, and Text using Recurrent Neural Networks
Figure 4 for Chains of Reasoning over Entities, Relations, and Text using Recurrent Neural Networks

Share this with someone who'll enjoy it:

Twitter IconFacebook IconLinkedin IconWhatsapp IconMessenger Icon

Our goal is to combine the rich multistep inference of symbolic logical reasoning with the generalization capabilities of neural networks. We are particularly interested in complex reasoning about entities and relations in text and large-scale knowledge bases (KBs). Neelakantan et al. (2015) use RNNs to compose the distributed semantics of multi-hop paths in KBs; however for multiple reasons, the approach lacks accuracy and practicality. This paper proposes three significant modeling advances: (1) we learn to jointly reason about relations, entities, and entity-types; (2) we use neural attention modeling to incorporate multiple paths; (3) we learn to share strength in a single RNN that represents logical composition across all relations. On a largescale Freebase+ClueWeb prediction task, we achieve 25% error reduction, and a 53% error reduction on sparse relations due to shared strength. On chains of reasoning in WordNet we reduce error in mean quantile by 84% versus previous state-of-the-art. The code and data are available at https://rajarshd.github.io/ChainsofReasoning

* accepted to EACL 2017 (fixed latex formatting in previous version)  
View paper onarxiv icon

Share this with someone who'll enjoy it:

Twitter IconFacebook IconLinkedin IconWhatsapp IconMessenger Icon