Judgment prediction for legal cases has attracted much research efforts forits practice use, of which the ultimate goal is prison term prediction. Whileexisting work merely predicts the total... | Continue reading
Wave-particle duality epitomizes the counterintuitive character of quantumphysics. A striking illustration is the quantum delay-choice experiment, whichis based on Wheeler's classic... | Continue reading
Multi-choice reading comprehension is a challenging task to select an answerfrom a set of candidate options when given passage and question. Previousapproaches usually only calculate... | Continue reading
Energy efficiency is one of the primary concerns when designing large scalecomputing systems. This makes reconfigurable hardware an attractive alternativeto load-store architectures, as it... | Continue reading
We introduce the dialog intent induction task and present a novel deepmulti-view clustering approach to tackle the problem. Dialog intent inductionaims at discovering user intents from user... | Continue reading
Gang-involved youth in cities such as Chicago sometimes post on social mediato express their aggression towards rival gangs and previous research hasdemonstrated that a deep learning approach... | Continue reading
Source code summarization -- creating natural language descriptions of sourcecode behavior -- is a rapidly-growing research topic with applications toautomatic documentation generation, program... | Continue reading
We present NarrativeTime, a new timeline-based annotation scheme for temporalorder of events in text, and a new densely annotated fiction corpus comparableto TimeBank-Dense. NarrativeTime is... | Continue reading
Feature norm datasets of human conceptual knowledge, collected in surveys ofhuman volunteers, yield highly interpretable models of word meaning and play animportant role in neurolinguistic... | Continue reading
Given a large amount of unannotated speech in a language with few resources,can we classify the speech utterances by topic? We show that this is possibleif text translations are available for... | Continue reading
We present a new explicit and stable numerical algorithm to solve thehomogeneous heat equation. We illustrate the performance of the new method inthe cases of two 2D systems with highly... | Continue reading
Recent predictions and experimental observations of high Tc superconductivityin hydrogen-rich materials at very high pressures are driving the search forsuperconductivity in the vicinity of room... | Continue reading
Learning target side syntactic structure has been shown to improve NeuralMachine Translation (NMT). However, incorporating syntax through latentvariables introduces additional complexity in... | Continue reading
Most existing work on adversarial data generation focuses on English. Forexample, PAWS (Paraphrase Adversaries from Word Scrambling) consists ofchallenging English paraphrase identification... | Continue reading
State-of-the-art semantic parsers rely on auto-regressive decoding, emittingone symbol at a time. When tested against complex databases that are unobservedat training time (zero-shot), the... | Continue reading
Coreference and many forms of ellipsis are similar to reading comprehensionquestions, in that in order to resolve these, we need to identify anappropriate text span in the previous discourse.... | Continue reading
Recent literature shows that large-scale language modeling provides excellentreusable sentence representations with both recurrent and self-attentivearchitectures. However, there has been less... | Continue reading
Advances in multicore processors and accelerators have opened the flood gatesto greater exploration and application of machine learning techniques to avariety of applications. These advances,... | Continue reading
Current research on hate speech analysis is typically oriented towardsmonolingual and single classification tasks. In this paper, we present a newmultilingual multi-aspect hate speech analysis... | Continue reading
We introduce and analyze a novel quantum machine learning model motivated byconvolutional neural networks. Our quantum convolutional neural network (QCNN)makes use of only $O(\log(N))$... | Continue reading
Recent researches prevalently used BiLSTM-CNN as a core module for NER in asequence-labeling setup. This paper formally shows the limitation of BiLSTM-CNNencoders in modeling cross-context... | Continue reading
Derivatives, mostly in the form of gradients and Hessians, are ubiquitous inmachine learning. Automatic differentiation (AD), also called algorithmicdifferentiation or simply "autodiff", is a... | Continue reading
Knowledge graphs typically undergo open-ended growth of new relations. Thiscannot be well handled by relation extraction that focuses on pre-definedrelations with sufficient training data. To... | Continue reading
We introduce a new classification task for scientific statements and releasea large-scale dataset for supervised learning. Our resource is derived from amachine-readable representation of the... | Continue reading
The requirements for many applications of state-of-the-art speech recognitionsystems include not only low word error rate (WER) but also low latency.Specifically, for many use-cases, the system... | Continue reading
PIE is a Prolog-embedded environment for automated reasoning on the basis offirst-order logic. It includes a versatile formula macro system and supportsthe creation of documents that intersperse... | Continue reading
KBSET supports a practical workflow for scholarly editing, based on usingLaTeX with dedicated commands for semantics-oriented markup and aProlog-implemented core system. Prolog plays there... | Continue reading
Recent years have seen great success in the use of neural seq2seq models onthe text-to-SQL task. However, little work has paid attention to how thesemodels generalize to realistic unseen data,... | Continue reading
Automatic Machine Learning (Auto-ML) has attracted more and more attention inrecent years, our work is to solve the problem of data drift, which means thatthe distribution of data will gradually... | Continue reading
This paper proposes a Transformer-based model to generate equations for mathword problems. It achieves much better results than RNN models when copy andalign mechanisms are not used, and can... | Continue reading
One of the biggest hurdles for customers when purchasing fashion online, isthe difficulty of finding products with the right fit. In order to provide abetter online shopping experience,... | Continue reading
Wikidata is one of the most edited knowledge bases which contains structureddata. It serves as the data source for many projects in the Wikimedia sphereand beyond. Since its inception in October... | Continue reading
Modern CPU architectures offer strong isolation guarantees towards userapplications in the form of enclaves. For instance, Intel's threat model forSGX assumes fully trusted enclaves, yet there... | Continue reading
In this work, we introduce a novel probabilistic representation of deeplearning, which provides an explicit explanation for the Deep Neural Networks(DNNs) in three aspects: (i) neurons define... | Continue reading
We use Very Long Baseline Interferometry to measure the proper motions ofthree black hole X-ray binaries (BHXBs). Using these results together with datafrom the literature and Gaia-DR2 to... | Continue reading
Developer contribution guidelines are used in social coding sites like GitHubto explain and shape the process a project expects contributors to follow. Theyset standards for all participants and... | Continue reading
Shallow syntax provides an approximation of phrase-syntactic structure ofsentences; it can be produced with high accuracy, and is computationally cheapto obtain. We investigate the role of... | Continue reading
Humans observe and interact with the world to acquire knowledge. However,most existing machine reading comprehension (MRC) tasks miss the interactive,information-seeking component of... | Continue reading
Image deraining is a fundamental, yet not well-solved problem in computervision and graphics. The traditional image deraining approaches commonly behaveineffectively in medium and heavy rain... | Continue reading
Existing machine reading comprehension (MRC) models do not scale effectivelyto real-world applications like web-level information retrieval and questionanswering (QA). We argue that this stems... | Continue reading
This paper describes neural-fortran, a parallel Fortran framework for neuralnetworks and deep learning. It features a simple interface to constructfeed-forward neural networks of arbitrary... | Continue reading
This paper presents a density-based topology optimization method fordesigning 3D thin-walled structures with adaptive meshing. Uniform wallthickness is achieved by simultaneously constraining... | Continue reading
Geotagging on social media has become an important proxy for understandingpeople's mobility and social events. Research that uses geotags to infer publicopinions relies on several key... | Continue reading
Fluid driven fractures propagate in the upper earth crust either naturally orin response to engineered fluid injections. The quantitative prediction oftheir evolution is critical in order to... | Continue reading
These notes gather recent results on robust statistical learning theory. Thegoal is to stress the main principles underlying the construction andtheoretical analysis of these estimators rather... | Continue reading
Trainable chatbots that exhibit fluent and human-like conversations remain abig challenge in artificial intelligence. Deep Reinforcement Learning (DRL) ispromising for addressing this challenge,... | Continue reading
Summary statistics such as the mean and variance are easily maintained forlarge, distributed data streams, but order statistics (i.e., sample quantiles)can only be approximately summarized.... | Continue reading
The deep supervised and reinforcement learning paradigms (among others) havethe potential to endow interactive multimodal social robots with the ability ofacquiring skills autonomously. But it... | Continue reading