• Publications
  • Influence
JFLEG: A Fluency Corpus and Benchmark for Grammatical Error Correction
TLDR
We present a new parallel corpus, JHU FLuency-Extended GUG corpus (JFLEG) for developing and evaluating grammatical error correction (GEC). Expand
  • 86
  • 20
  • PDF
Ground Truth for Grammaticality Correction Metrics
TLDR
We establish a ground truth for GEC by conducting a human evaluation and producing a human ranking of the systems entered into the CoNLL-2014 Shared Task on GEC. Expand
  • 61
  • 20
  • PDF
WINOGRANDE: An Adversarial Winograd Schema Challenge at Scale
TLDR
We introduce WinoGrande, a large-scale dataset of 44k problems, inspired by the original WSC design, but adjusted to improve both the scale and the hardness of the dataset. Expand
  • 86
  • 16
  • PDF
Abductive Commonsense Reasoning
TLDR
In this paper, we present the first study that investigates the viability of language-based abductive reasoning. Expand
  • 65
  • 8
  • PDF
Robsut Wrod Reocginiton via Semi-Character Recurrent Neural Network
TLDR
We propose a word recognition model based on a semi-character level recurrent neural network (scRNN). Expand
  • 50
  • 8
  • PDF
Grammatical Error Correction with Neural Reinforcement Learning
TLDR
We propose a neural encoder-decoder model with reinforcement learning for grammatical error correction (GEC). Expand
  • 44
  • 8
  • PDF
Reassessing the Goals of Grammatical Error Correction: Fluency Instead of Grammaticality
TLDR
The field of grammatical error correction (GEC) has grown substantially in recent years, with research directed at both evaluation metrics and improved system performance against those metrics. Expand
  • 37
  • 7
  • PDF
There's No Comparison: Reference-less Evaluation Metrics in Grammatical Error Correction
TLDR
We show that reference-less grammaticality metrics correlate very strongly with human judgments and are competitive with the leading reference-based evaluation metrics. Expand
  • 31
  • 7
  • PDF
Universal Decompositional Semantics on Universal Dependencies
TLDR
This paper describes the Universal Decompositional Semantics (Decomp) project, which aims to augments Universal Dependencies (UD) data sets with robust, scalable semantic annotations based in linguistic theory. Expand
  • 81
  • 6
  • PDF
GLEU Without Tuning
TLDR
The GLEU metric was proposed for evaluating grammatical error corrections using n-gram overlap with a set of reference sentences, as opposed to precision/recall. Expand
  • 19
  • 5
  • PDF