• Publications
  • Influence
Frame-Semantic Parsing
TLDR
We solve the problem of frame-semantic parsing using a two-stage statistical model that takes lexical targets (i.e., content words and phrases) in their sentential contexts and predicts frame-Semantic structures. Expand
  • 241
  • 44
  • PDF
Turning on the Turbo: Fast Third-Order Non-Projective Turbo Parsers
TLDR
We present fast, accurate, direct nonprojective dependency parsers with thirdorder features. Expand
  • 186
  • 32
  • PDF
From Softmax to Sparsemax: A Sparse Model of Attention and Multi-Label Classification
TLDR
We propose sparsemax, a new activation function similar to the traditional softmax, but able to output sparse probabilities. Expand
  • 224
  • 27
  • PDF
An Augmented Lagrangian Approach to Constrained MAP Inference
TLDR
We propose a new algorithm for approximate MAP inference on factor graphs, by combining augmented Lagrangian optimization with the dual decomposition method. Expand
  • 124
  • 20
  • PDF
Marian: Fast Neural Machine Translation in C++
TLDR
We present Marian, an efficient and self-contained Neural Machine Translation framework with an integrated automatic differentiation engine based on dynamic computation graphs. Expand
  • 178
  • 19
  • PDF
Concise Integer Linear Programming Formulations for Dependency Parsing
TLDR
We formulate the problem of non-projective dependency parsing as a polynomial-sized integer linear program. Expand
  • 180
  • 14
  • PDF
Dual Decomposition with Many Overlapping Components
TLDR
We solve an LP relaxation through a recently proposed consensus al- gorithm, DD-ADMM, which is suitable for problems with many overlapping components. Expand
  • 79
  • 14
  • PDF
Turbo Parsers: Dependency Parsing by Approximate Variational Inference
TLDR
We present a unified view of two state-of-the-art non-projective dependency parsers, both approximate: the loopy belief propagation parser of Smith and Eisner (2008) and the relaxed linear program of Martins et al. (2009). Expand
  • 135
  • 13
  • PDF
Stacking Dependency Parsers
TLDR
We explore a stacked framework for learning to predict dependency structures for natural language sentences through the use of stacked learning (Wolpert, 1992). Expand
  • 100
  • 12
  • PDF
AD3: alternating directions dual decomposition for MAP inference in graphical models
TLDR
We present AD3, a new algorithm for approximate maximum a posteriori (MAP) inference on factor graphs, based on the alternating directions method of multipliers. Expand
  • 48
  • 12
  • PDF