October 15, 2019

2164 words 11 mins read

Paper Group NANR 90

Paper Group NANR 90

SumeCzech: Large Czech News-Based Summarization Dataset. HybridNet: A Hybrid Neural Architecture to Speed-up Autoregressive Models. Using Paraphrasing and Memory-Augmented Models to Combat Data Sparsity in Question Interpretation with a Virtual Patient Dialogue System. Creative Language Encoding under Censorship. ENRICHMENT OF FEATURES FOR CLASSIFI …

SumeCzech: Large Czech News-Based Summarization Dataset

Title SumeCzech: Large Czech News-Based Summarization Dataset
Authors Milan Straka, Nikita Mediankin, Tom Kocmi, Zden{\v{e}}k {\v{Z}}abokrtsk{'y}, Vojt{\v{e}}ch Hude{\v{c}}ek, Jan Haji{\v{c}}
Abstract
Tasks Document Summarization, Machine Translation, Sentence Compression
Published 2018-05-01
URL https://www.aclweb.org/anthology/L18-1551/
PDF https://www.aclweb.org/anthology/L18-1551
PWC https://paperswithcode.com/paper/sumeczech-large-czech-news-based
Repo
Framework

HybridNet: A Hybrid Neural Architecture to Speed-up Autoregressive Models

Title HybridNet: A Hybrid Neural Architecture to Speed-up Autoregressive Models
Authors Yanqi Zhou, Wei Ping, Sercan Arik, Kainan Peng, Greg Diamos
Abstract This paper introduces HybridNet, a hybrid neural network to speed-up autoregressive models for raw audio waveform generation. As an example, we propose a hybrid model that combines an autoregressive network named WaveNet and a conventional LSTM model to address speech synthesis. Instead of generating one sample per time-step, the proposed HybridNet generates multiple samples per time-step by exploiting the long-term memory utilization property of LSTMs. In the evaluation, when applied to text-to-speech, HybridNet yields state-of-art performance. HybridNet achieves a 3.83 subjective 5-scale mean opinion score on US English, largely outperforming the same size WaveNet in terms of naturalness and provide 2x speed up at inference.
Tasks Speech Synthesis
Published 2018-01-01
URL https://openreview.net/forum?id=rJoXrxZAZ
PDF https://openreview.net/pdf?id=rJoXrxZAZ
PWC https://paperswithcode.com/paper/hybridnet-a-hybrid-neural-architecture-to
Repo
Framework

Using Paraphrasing and Memory-Augmented Models to Combat Data Sparsity in Question Interpretation with a Virtual Patient Dialogue System

Title Using Paraphrasing and Memory-Augmented Models to Combat Data Sparsity in Question Interpretation with a Virtual Patient Dialogue System
Authors Lifeng Jin, David King, Amad Hussein, Michael White, Douglas Danforth
Abstract When interpreting questions in a virtual patient dialogue system one must inevitably tackle the challenge of a long tail of relatively infrequently asked questions. To make progress on this challenge, we investigate the use of paraphrasing for data augmentation and neural memory-based classification, finding that the two methods work best in combination. In particular, we find that the neural memory-based approach not only outperforms a straight CNN classifier on low frequency questions, but also takes better advantage of the augmented data created by paraphrasing, together yielding a nearly 10{%} absolute improvement in accuracy on the least frequently asked questions.
Tasks Data Augmentation, One-Shot Learning
Published 2018-06-01
URL https://www.aclweb.org/anthology/W18-0502/
PDF https://www.aclweb.org/anthology/W18-0502
PWC https://paperswithcode.com/paper/using-paraphrasing-and-memory-augmented
Repo
Framework

Creative Language Encoding under Censorship

Title Creative Language Encoding under Censorship
Authors Heng Ji, Kevin Knight
Abstract People often create obfuscated language for online communication to avoid Internet censorship, share sensitive information, express strong sentiment or emotion, plan for secret actions, trade illegal products, or simply hold interesting conversations. In this position paper we systematically categorize human-created obfuscated language on various levels, investigate their basic mechanisms, give an overview on automated techniques needed to simulate human encoding. These encoders have potential to frustrate and evade, co-evolve with dynamic human or automated decoders, and produce interesting and adoptable code words. We also summarize remaining challenges for future research on the interaction between Natural Language Processing (NLP) and encryption, and leveraging NLP techniques for encoding and decoding.
Tasks
Published 2018-08-01
URL https://www.aclweb.org/anthology/W18-4203/
PDF https://www.aclweb.org/anthology/W18-4203
PWC https://paperswithcode.com/paper/creative-language-encoding-under-censorship
Repo
Framework

ENRICHMENT OF FEATURES FOR CLASSIFICATION USING AN OPTIMIZED LINEAR/NON-LINEAR COMBINATION OF INPUT FEATURES

Title ENRICHMENT OF FEATURES FOR CLASSIFICATION USING AN OPTIMIZED LINEAR/NON-LINEAR COMBINATION OF INPUT FEATURES
Authors Mehran Taghipour-Gorjikolaie, Seyyed Mohammad Razavi, Javad Sadri
Abstract Automatic classification of objects is one of the most important tasks in engineering and data mining applications. Although using more complex and advanced classifiers can help to improve the accuracy of classification systems, it can be done by analyzing data sets and their features for a particular problem. Feature combination is the one which can improve the quality of the features. In this paper, a structure similar to Feed-Forward Neural Network (FFNN) is used to generate an optimized linear or non-linear combination of features for classification. Genetic Algorithm (GA) is applied to update weights and biases. Since nature of data sets and their features impact on the effectiveness of combination and classification system, linear and non-linear activation functions (or transfer function) are used to achieve more reliable system. Experiments of several UCI data sets and using minimum distance classifier as a simple classifier indicate that proposed linear and non-linear intelligent FFNN-based feature combination can present more reliable and promising results. By using such a feature combination method, there is no need to use more powerful and complex classifier anymore.
Tasks
Published 2018-01-01
URL https://openreview.net/forum?id=HJqUtdOaZ
PDF https://openreview.net/pdf?id=HJqUtdOaZ
PWC https://paperswithcode.com/paper/enrichment-of-features-for-classification
Repo
Framework

Reconstructing Thin Structures of Manifold Surfaces by Integrating Spatial Curves

Title Reconstructing Thin Structures of Manifold Surfaces by Integrating Spatial Curves
Authors Shiwei Li, Yao Yao, Tian Fang, Long Quan
Abstract The manifold surface reconstruction in multi-view stereo often fails in retaining thin structures due to incomplete and noisy reconstructed point clouds. In this paper, we address this problem by leveraging spatial curves. The curve representation in nature is advantageous in modeling thin and elongated structures, implying topology and connectivity information of the underlying geometry, which exactly compensates the weakness of scattered point clouds. We present a novel surface reconstruction method using both curves and point clouds. First, we propose a 3D curve reconstruction algorithm based on the initialize-optimize-expand strategy. Then, tetrahedra are constructed from points and curves, where the volumes of thin structures are robustly preserved by the Curve-conformed Delaunay Refinement. Finally, the mesh surface is extracted from tetrahedra by a graph optimization. The method has been intensively evaluated on both synthetic and real-world datasets, showing significant improvements over state-of-the-art methods.
Tasks
Published 2018-06-01
URL http://openaccess.thecvf.com/content_cvpr_2018/html/Li_Reconstructing_Thin_Structures_CVPR_2018_paper.html
PDF http://openaccess.thecvf.com/content_cvpr_2018/papers/Li_Reconstructing_Thin_Structures_CVPR_2018_paper.pdf
PWC https://paperswithcode.com/paper/reconstructing-thin-structures-of-manifold
Repo
Framework

Generative Discovery of Relational Medical Entity Pairs

Title Generative Discovery of Relational Medical Entity Pairs
Authors Chenwei Zhang, Yaliang Li, Nan Du, Wei Fan, Philip S. Yu
Abstract Online healthcare services can provide the general public with ubiquitous access to medical knowledge and reduce the information access cost for both individuals and societies. To promote these benefits, it is desired to effectively expand the scale of high-quality yet novel relational medical entity pairs that embody rich medical knowledge in a structured form. To fulfill this goal, we introduce a generative model called Conditional Relationship Variational Autoencoder (CRVAE), which can discover meaningful and novel relational medical entity pairs without the requirement of additional external knowledge. Rather than discriminatively identifying the relationship between two given medical entities in a free-text corpus, we directly model and understand medical relationships from diversely expressed medical entity pairs. The proposed model introduces the generative modeling capacity of variational autoencoder to entity pairs, and has the ability to discover new relational medical entity pairs solely based on the existing entity pairs. Beside entity pairs, relationship-enhanced entity representations are obtained as another appealing benefit of the proposed method. Both quantitative and qualitative evaluations on real-world medical datasets demonstrate the effectiveness of the proposed method in generating relational medical entity pairs that are meaningful and novel.
Tasks
Published 2018-01-01
URL https://openreview.net/forum?id=BJhxcGZCW
PDF https://openreview.net/pdf?id=BJhxcGZCW
PWC https://paperswithcode.com/paper/generative-discovery-of-relational-medical
Repo
Framework

Generating Questions for Reading Comprehension using Coherence Relations

Title Generating Questions for Reading Comprehension using Coherence Relations
Authors Takshak Desai, Parag Dakle, Dan Moldovan
Abstract In this paper, we have proposed a technique for generating complex reading comprehension questions from a discourse that are more useful than factual ones derived from assertions. Our system produces a set of general-level questions using coherence relations and a set of well-defined syntactic transformations on the input text. Generated questions evaluate comprehension abilities like a comprehensive analysis of the text and its structure, correct identification of the author{'}s intent, a thorough evaluation of stated arguments; and a deduction of the high-level semantic relations that hold between text spans. Experiments performed on the RST-DT corpus allow us to conclude that our system possesses a strong aptitude for generating intricate questions. These questions are capable of effectively assessing a student{'}s interpretation of the text.
Tasks Reading Comprehension
Published 2018-07-01
URL https://www.aclweb.org/anthology/W18-3701/
PDF https://www.aclweb.org/anthology/W18-3701
PWC https://paperswithcode.com/paper/generating-questions-for-reading
Repo
Framework

3D Scene Flow from 4D Light Field Gradients

Title 3D Scene Flow from 4D Light Field Gradients
Authors Sizhuo Ma, Brandon M. Smith, Mohit Gupta
Abstract This paper presents novel techniques for recovering 3D dense scene flow, based on differential analysis of 4D light fields. The key enabling result is a per-ray linear equation, called the ray flow equation, that relates 3D scene flow to 4D light field gradients. The ray flow equation is invariant to 3D scene structure and applicable to a general class of scenes, but is underconstrained (3 unknowns per equation). Thus, additional constraints must be imposed to recover motion. We develop two families of scene flow algorithms by leveraging the structural similarity between ray flow and optical flow equations: local “Lucas-Kanade” ray flow and global “Horn-Schunck” ray flow, inspired by corresponding optical flow methods. We also develop a combined local-global method by utilizing the correspondence structure in the light fields. We demonstrate high precision 3D scene flow recovery for a wide range of scenarios, including rotation and non-rigid motion. We analyze the theoretical and practical performance limits of the proposed techniques via the light field structure tensor, a 3x3 matrix that encodes the local structure of light fields. We envision that the proposed analysis and algorithms will lead to design of future light-field cameras that are optimized for motion sensing, in addition to depth sensing.
Tasks Optical Flow Estimation
Published 2018-09-01
URL http://openaccess.thecvf.com/content_ECCV_2018/html/Sizhuo_3D_Motion_Sensing_ECCV_2018_paper.html
PDF http://openaccess.thecvf.com/content_ECCV_2018/papers/Sizhuo_3D_Motion_Sensing_ECCV_2018_paper.pdf
PWC https://paperswithcode.com/paper/3d-scene-flow-from-4d-light-field-gradients
Repo
Framework

SRFeat: Single Image Super-Resolution with Feature Discrimination

Title SRFeat: Single Image Super-Resolution with Feature Discrimination
Authors Seong-Jin Park, Hyeongseok Son, Sunghyun Cho, Ki-Sang Hong, Seungyong Lee
Abstract Generative adversarial networks (GANs) have recently been adopted to single image super resolution (SISR) and showed impressive results with realistically synthesized high-frequency textures. However, the results of such GAN based approaches tend to include less meaningful high-frequency noise that is irrelevant to the input image. In this paper, we propose a novel GAN-based SISR method that overcomes the limitation and produces more realistic results by attaching an additional discriminator that works in the feature domain. Our additional discriminator encourages the generator to produce structural high-frequency features rather than noisy artifacts as it distinguishes synthetic and real images in terms of features. We also design a new generator that utilizes long-range skip connections so that information between distant layers can be transferred more effectively. Experiments show that our method achieves the state-of-the-art performance in terms of both PSNR and perceptual quality compared to recent GAN-based methods.
Tasks Image Super-Resolution, Super-Resolution
Published 2018-09-01
URL http://openaccess.thecvf.com/content_ECCV_2018/html/Seong-Jin_Park_SRFeat_Single_Image_ECCV_2018_paper.html
PDF http://openaccess.thecvf.com/content_ECCV_2018/papers/Seong-Jin_Park_SRFeat_Single_Image_ECCV_2018_paper.pdf
PWC https://paperswithcode.com/paper/srfeat-single-image-super-resolution-with
Repo
Framework

A Comparative Study of Embedding Models in Predicting the Compositionality of Multiword Expressions

Title A Comparative Study of Embedding Models in Predicting the Compositionality of Multiword Expressions
Authors N, Navnita akumar, Bahar Salehi, Timothy Baldwin
Abstract In this paper, we perform a comparative evaluation of off-the-shelf embedding models over the task of compositionality prediction of multiword expressions(``MWEs’'). Our experimental results suggest that character- and document-level models capture knowledge of MWE compositionality and are effective in modelling varying levels of compositionality, with the advantage over word-level models that they do not require token-level identification of MWEs in the training corpus. |
Tasks Information Retrieval, Representation Learning, Word Embeddings
Published 2018-12-01
URL https://www.aclweb.org/anthology/U18-1009/
PDF https://www.aclweb.org/anthology/U18-1009
PWC https://paperswithcode.com/paper/a-comparative-study-of-embedding-models-in
Repo
Framework

YNU Deep at SemEval-2018 Task 12: A BiLSTM Model with Neural Attention for Argument Reasoning Comprehension

Title YNU Deep at SemEval-2018 Task 12: A BiLSTM Model with Neural Attention for Argument Reasoning Comprehension
Authors Peng Ding, Xiaobing Zhou
Abstract This paper describes the system submitted to SemEval-2018 Task 12 (The Argument Reasoning Comprehension Task). Enabling a computer to understand a text so that it can answer comprehension questions is still a challenging goal of NLP. We propose a Bidirectional LSTM (BiLSTM) model that reads two sentences separated by a delimiter to determine which warrant is correct. We extend this model with a neural attention mechanism that encourages the model to make reasoning over the given claims and reasons. Officially released results show that our system ranks 6th among 22 submissions to this task.
Tasks Constituency Parsing, Language Modelling, Machine Translation, Reading Comprehension
Published 2018-06-01
URL https://www.aclweb.org/anthology/S18-1189/
PDF https://www.aclweb.org/anthology/S18-1189
PWC https://paperswithcode.com/paper/ynu-deep-at-semeval-2018-task-12-a-bilstm
Repo
Framework

New Baseline in Automatic Speech Recognition for Northern S'ami

Title New Baseline in Automatic Speech Recognition for Northern S'ami
Authors Juho Leinonen, Peter Smit, S{'a}mi Virpioja, Mikko Kurimo
Abstract
Tasks Language Modelling, Speech Recognition
Published 2018-01-01
URL https://www.aclweb.org/anthology/W18-0208/
PDF https://www.aclweb.org/anthology/W18-0208
PWC https://paperswithcode.com/paper/new-baseline-in-automatic-speech-recognition
Repo
Framework

Extracting inflectional class assignment in Pite Saami: Nouns, verbs and those pesky adjectives

Title Extracting inflectional class assignment in Pite Saami: Nouns, verbs and those pesky adjectives
Authors Joshua Wilbur
Abstract
Tasks
Published 2018-01-01
URL https://www.aclweb.org/anthology/W18-0213/
PDF https://www.aclweb.org/anthology/W18-0213
PWC https://paperswithcode.com/paper/extracting-inflectional-class-assignment-in
Repo
Framework

Statistical Learning Theory and Linguistic Typology: a Learnability Perspective on OT’s Strict Domination

Title Statistical Learning Theory and Linguistic Typology: a Learnability Perspective on OT’s Strict Domination
Authors {'E}mile Enguehard, Edward Flemming, Giorgio Magri
Abstract
Tasks
Published 2018-01-01
URL https://www.aclweb.org/anthology/W18-0301/
PDF https://www.aclweb.org/anthology/W18-0301
PWC https://paperswithcode.com/paper/statistical-learning-theory-and-linguistic
Repo
Framework
comments powered by Disqus