• Publications
  • Influence
Select, Answer and Explain: Interpretable Multi-hop Reading Comprehension over Multiple Documents
TLDR
We propose an effective and interpretable Select, Answer and Explain (SAE) system to solve the multi-document RC problem. Expand
  • 23
  • 4
  • PDF
Multi-hop Reading Comprehension across Multiple Documents by Reasoning over Heterogeneous Graphs
TLDR
We introduce a heterogeneous graph with different types of nodes and edges, which is named as Heterogeneous Document-Entity (HDE) graph. Expand
  • 30
  • 3
  • PDF
Online speaking rate estimation using recurrent neural networks
TLDR
We propose an online speaking rate estimation model based on recurrent neural networks based on a set of speech features that are known to correlate with speech rhythm. Expand
  • 11
  • 2
Detection of long-term changes in precipitation and discharge in the Meuse basin
Long records (>1911) of discharge and precipitation in the Meuse basin have been investigated by statistical methods for detection of non-homogeneity (trends and jumps) in the data series. Over theExpand
  • 13
  • 2
Speech enhancement based on Deep Neural Networks with skip connections
Speech enhancement under noise condition has always been an intriguing research topic. In this paper, we propose a new Deep Neural Networks (DNNs) based architecture for speech enhancement. InExpand
  • 23
  • 1
  • PDF
Convex Weighting Criteria for Speaking Rate Estimation
TLDR
Speaking rate estimation directly from the speech waveform is a long-standing problem in speech signal processing. Expand
  • 21
  • 1
  • PDF
Validating Prediction Models of Kidney Transplant Outcome Using Single Center Data
Prediction of kidney transplant outcome represents an important and clinically relevant problem. Although several prediction models have been proposed based on large, national collections of data,Expand
  • 10
  • 1
Models for objective evaluation of dysarthric speech from data annotated by multiple listeners
TLDR
We propose a new algorithm to solve the multi-annotator problem for regression-based objective evaluation of dysarthric speech and show that our method outperforms other similar approaches. Expand
  • 5
  • 1
Multiple instance learning with graph neural networks
TLDR
We propose a new end-to-end graph neural network (GNN) based algorithm for MIL: we treat each bag as a graph and use GNN to learn the bag embedding in order to explore the useful structural information among instances in bags. Expand
  • 9
  • 1
  • PDF
Improving voice quality of HMM-based speech synthesis using voice conversion method
TLDR
We propose to use voice conversion method to transform synthetic speech toward the original so as to improve its quality. Expand
  • 3
  • 1
  • PDF