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

July 05, 2016 · Entered Twilight · 🏛 Conference of the European Chapter of the Association for Computational Linguistics

"No code URL or promise found in abstract"
"Derived repo from GitHub Pages (backfill)"

Evidence collected by the PWNC Scanner

Repo contents: README.md, data, eval, examples, model, run_scripts, util, vocab, wordnet_experiment

Authors Rajarshi Das, Arvind Neelakantan, David Belanger, Andrew McCallum arXiv ID 1607.01426 Category cs.CL: Computation & Language Citations 274 Venue Conference of the European Chapter of the Association for Computational Linguistics Repository https://github.com/rajarshd/ChainsofReasoning ⭐ 86 Last Checked 7 days ago

Abstract

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