Programmers typically organize executable source code using high-level codingpatterns or idiomatic structures such as nested loops, exception handlers andrecursive blocks, rather than as... | Continue reading
Kingdomino is a board game designed by Bruno Cathala and edited by BlueOrange since 2016. The goal is to place $2 \times 1$ dominoes on a grid layout,and get a better score than other players.... | Continue reading
Neural models for NLP typically use large numbers of parameters to reachstate-of-the-art performance, which can lead to excessive memory usage andincreased runtime. We present a structure... | Continue reading
While a lot of progress has been made in recent years, the dynamics oflearning in deep nonlinear neural networks remain to this day largelymisunderstood. In this work, we study the case of... | Continue reading
Several methods of estimating the mutual information of random variables havebeen developed in recent years. They can prove valuable for novel approaches tolearning statistically independent... | Continue reading
When an entity name contains other names within it, the identification of allcombinations of names can become difficult and expensive. We propose a newmethod to recognize not only outermost... | Continue reading
Core to the vision-and-language navigation (VLN) challenge is building robustinstruction representations and action decoding schemes, which can generalizewell to previously unseen instructions... | Continue reading
Treebank translation is a promising method for cross-lingual transfer ofsyntactic dependency knowledge. The basic idea is to map dependency arcs from asource treebank to its target translation... | Continue reading
Recent studies have shown that word embeddings exhibit gender bias inheritedfrom the training corpora. However, most studies to date have focused onquantifying and mitigating such bias only in... | Continue reading
Current state-of-the-art neural machine translation (NMT) uses a deepmulti-head self-attention network with no explicit phrase information. However,prior work on statistical machine translation... | Continue reading
The latest work on language representations carefully integratescontextualized features into language model training, which enables a series ofsuccess especially in various machine reading... | Continue reading
Multilingual Neural Machine Translation (NMT) models have yielded largeempirical success in transfer learning settings. However, these black-boxrepresentations are poorly understood, and their... | Continue reading
Diplomacy is a seven-player non-stochastic, non-cooperative game, whereagents acquire resources through a mix of teamwork and betrayal. Reliance ontrust and coordination makes Diplomacy the... | Continue reading
Power spectral densities are a common, convenient, and powerful way toanalyze signals. So much so that they are now broadly deployed across thesciences and engineering---from quantum physics to... | Continue reading
To capture the systemic complexity of international financial systems,network data is an important prerequisite. However, dyadic data is often notavailable, raising the need for methods that... | Continue reading
Let's Plays of video games represent a relatively unexplored area forexperimental AI in games. In this short paper, we discuss an approach togenerate automated commentary for Let's Play videos,... | Continue reading
While deep neural models have gained successes on information extractiontasks, they become less reliable when the amount of labeled data is limited. Inthis paper, we study relation extraction... | Continue reading
Commonsense reasoning aims to empower machines with the human ability to makepresumptions about ordinary situations in our daily life. In this paper, wepropose a textual inference framework for... | Continue reading
Despite of the recent success of collective entity linking (EL) methods,these "global" inference methods may yield sub-optimal results when the"all-mention coherence" assumption breaks, and... | Continue reading
Modern deep learning-based recommendation systems exploit hundreds tothousands of different categorical features, each with millions of differentcategories ranging from clicks to posts. To... | Continue reading
The state of the art in machine translation (MT) is governed by neuralapproaches, which typically provide superior translation accuracy overstatistical approaches. However, on the closely... | Continue reading
Abstractive summarization systems aim to produce more coherent and concisesummaries than their extractive counterparts. Popular neural models haveachieved impressive results for single-document... | Continue reading
This paper presents a new metric called TIGEr for the automatic evaluation ofimage captioning systems. Popular metrics, such as BLEU and CIDEr, are basedsolely on text matching between reference... | Continue reading
Task-oriented dialog systems need to know when a query falls outside theirrange of supported intents, but current text classification corpora only definelabel sets that cover every example. We... | Continue reading
We describe a detailed analysis of a sample of large benchmark of commonsensereasoning problems that has been automatically obtained from WordNet, SUMO andtheir mapping. The objective is to... | Continue reading
We present PaLM, a hybrid parser and neural language model. Building on anRNN language model, PaLM adds an attention layer over text spans in the leftcontext. An unsupervised constituency parser... | Continue reading
Given access to accurate solutions of the many-electron Schrödingerequation, nearly all chemistry could be derived from first principles. Exactwavefunctions of interesting chemical systems are... | Continue reading
Language models are generally trained on data spanning a wide range of topics(e.g., news, reviews, fiction), but they might be applied to an a prioriunknown target distribution (e.g., restaurant... | Continue reading
In this paper, we explore new approaches to combining information encodedwithin the learned representations of auto-encoders. We explore models that arecapable of combining the attributes of... | Continue reading
Answer Set Programming (ASP) has become, the paradigm of choice in the fieldof logic programming and non-monotonic reasoning. Thanks to the availability ofefficient solvers, ASP has been... | Continue reading
Bilinear diagonal models for knowledge graph embedding (KGE), such asDistMult and ComplEx, balance expressiveness and computational efficiency byrepresenting relations as diagonal matrices.... | Continue reading
Gauge fields, real or synthetic, are crucial for understanding andmanipulation of physical systems. The associated geometric phases can bemeasured, for example, from the Aharonov--Bohm... | Continue reading
This paper is the preprint of an invited commentary on Lake et al'sBehavioral and Brain Sciences article titled "Building machines that learn andthink like people". Lake et al's paper offers a... | Continue reading
Link prediction is one of the fundamental problems in social networkanalysis. A common set of techniques for link prediction rely on similaritymetrics which use the topology of the observed... | Continue reading
This thesis focuses on process mining on event data where such a normativespecification is absent and, as a result, the event data is less structured.The thesis puts special emphasis on one... | Continue reading
Most machine translation systems generate text autoregressively from left toright. We, instead, use a masked language modeling objective to train a modelto predict any subset of the target... | Continue reading
AI has achieved remarkable mastery over games such as Chess, Go, and Poker,and even Jeopardy, but the rich variety of standardized exams has remained alandmark challenge. Even in 2016, the best... | Continue reading
This paper introduces R2D3, an agent that makes efficient use ofdemonstrations to solve hard exploration problems in partially observableenvironments with highly variable initial conditions. We... | Continue reading
The strength of obfuscated software has increased over the recent years.Compiler based obfuscation has become the de facto standard in the industry andrecent papers also show that injection of... | Continue reading
Generative models have achieved impressive results in many domains includingimage and text generation. In the natural sciences, generative models have ledto rapid progress in automated drug... | Continue reading
Query complexity is a common tool for comparing quantum and classicalcomputation, and it has produced many examples of how quantum algorithms differfrom classical ones. Here we investigate in... | Continue reading
We propose weakly supervised language localization networks (WSLLN) to detectevents in long, untrimmed videos given language queries. To learn thecorrespondence between visual segments and... | Continue reading
We focus on the cross-domain context-dependent text-to-SQL generation task.Based on the observation that adjacent natural language questions are oftenlinguistically dependent and their... | Continue reading
Deep learning systems thrive on abundance of labeled training data but suchdata is not always available, calling for alternative methods of supervision.One such method is expectation... | Continue reading
Program synthesis of general-purpose source code from natural languagespecifications is challenging due to the need to reason about high-levelpatterns in the target program and low-level... | Continue reading
Transition-based top-down parsing with pointer networks has achievedstate-of-the-art results in multiple parsing tasks, while having a linear timecomplexity. However, the decoder of these... | Continue reading
Dialogue management (DM) plays a key role in the quality of the interactionwith the user in a task-oriented dialogue system. In most existing approaches,the agent predicts only one DM policy... | Continue reading
Both syntactic and semantic structures are key linguistic contextual clues,in which parsing the latter has been well shown beneficial from parsing theformer. However, few works ever made an... | Continue reading