TY - JOUR
AU - 
AB - Strong Baselines for Simple Question Answering over Knowledge Graphs with and without Neural Networks Salman Mohammed, Peng Shi, and Jimmy Lin David R. Cheriton School of Computer Science University of Waterloo smohammed1993@gmail.com, {peng.shi,jimmylin}@uwaterloo.ca Abstract architectures, when properly tuned, outperform some more recent models; Vaswani et al. (2017) We examine the problem of question answer- showed that the dominant approach to sequence ing over knowledge graphs, focusing on sim- transduction using complex encoder–decoder net- ple questions that can be answered by the works with attention mechanisms work just as lookup of a single fact. Adopting a straight- well with the attention module only, yielding net- forward decomposition of the problem into en- works that are far simpler and easier to train. tity detection, entity linking, relation predic- tion, and evidence combination, we explore In line with an emerging thread of research that simple yet strong baselines. On the popular aims to improve empirical rigor in our field by fo- SIMPLEQUESTIONS dataset, we find that ba- cusing on knowledge and insights, as opposed to sic LSTMs and GRUs plus a few heuristics simply “winning” (Sculley et al., 2018), we take yield accuracies that approach the state of the the approach of 
TI -  Strong Baselines for Simple Question Answering over Knowledge Graphs with and without Neural Networks 
JF - Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers)
DO - 10.18653/v1/n18-2047
DA - 2018-01-01
UR - https://www.deepdyve.com/lp/unpaywall/strong-baselines-for-simple-question-answering-over-knowledge-graphs-6OBepWez4j
DP - DeepDyve
ER -