Research from Samsung Research will be presented in the oral and poster sessions of the 2018 Conference on Empirical Methods in Natural Language Processing (EMNLP 2018). One of the top-tier conferences in natural language processing, EMNLP is organized by the Association for Computational Linguistics (ACL). First held in 1996, this year will be its 23rd and will be held in Brussels, Belgium from November 2 to 4, 2018. Last year, the acceptance rates of long and short papers were 26% and 18%, respectively.
"Memo Reader: Large-Scale Reading Comprehension through Neural Memory Controller"
Seohyun Back, Seunghak Yu, Sathish Indurthi, Jihie Kim and Jaegul Choo
This paper proposes a novel memory-augmented deep neural network architecture with skip connections to handle long-range dependency in reading comprehension tasks, as well as an expanded GRU architecture with dense connections that compensate potential information distortion occurring in the memory.
"Cut to the Chase: A Context Zoom-In Network for Reading Comprehension"
Sathish Indurthi, Seunghak Yu, Seohyun Back and Heriberto Cuayáhuitl
This paper presents a novel neural-based architecture that can extract relevant regions, based on a given question-document pair and subsequently generate a well-formed answer.
"Supervised Clustering of Questions into Intents for Dialog System Applications"
Iryna Haponchyk, Antonio Uva, Seunghak Yu, Olga Uryupina and Alessandro Moschitti
This paper proposes a model for automatic clustering of questions into user intents using semantic classifiers, along with a novel idea of supervised clustering methods based on a structured output.