We describe our approach to create and deliver a custom voice for aconversational AI use-case . Our proposed cloudarchitecture enables for fast voice delivery, making it possible to talk to the digital version of Albert Einstein in real-time . We use a custom dictionary for selected words to ensure their proper pronunciation .…

## Fine Grained Causality Extraction From Natural Language Requirements Using Recursive Neural Tensor Networks

Causal relations (e.g., If A, then B) are prevalent in functional requirements . For various applications of AI4RE, automatically extracting such causal statements are a basic necessity . We lack an approach that is able to extract causal relations from natural language requirements in fine-grainedform .…

## How to Tell Deep Neural Networks What We Know

We present a short survey of ways in which existing scientific knowledge are included when constructing models with neural networks . The inclusion of domain-knowledge is of special interest not just to constructing scientificassistants, but also, many other areas that involve understanding data using human-machine collaboration .…

## Strategic Mitigation of Agent Inattention in Drivers with Open Quantum Cognition Models

State-of-the-art driver-assist systems have failed to effectively mitigatedriver inattention and had minimal impacts on the ever-growing number of roadmishaps . This is because traditional human-machine interaction settings are modeled in classical and behavioralgame-theoretic domains . We propose a novel equilibrium notion in human-systeminteraction games, where the system maximizes its expected utility and humandecisions can be characterized using any general decision model .…

## How Do Pedophiles Tweet Investigating the Writing Styles and Online Personas of Child Cybersex Traffickers in the Philippines

Child sex peddlers spread illegal content and target minors for sexual activities on Twitter in the Philippines using Natural Language Processing techniques . Results of our study show frequently used and co-occurring words that traffickers use tospread content . Four main roles played by these entities that contribute to the proliferation of child pornography in the country .…

## Extension of additive valuations to general valuations on the existence of EFX

Envy-freeness is one of the most widely studied notions in fair division . Since envy-free allocations do not always exist when items are indivisible . We show that an EFXallocation always exists (i) when all agents have one of two general valuationsor (ii) when the number of items is at most $n+3$ .…

## CATE CAusality Tree Extractor from Natural Language Requirements

Causal relations (If A, then B) are prevalent in requirements artifacts . We lack an approach capable of extracting causal relations from natural language with reasonable performance . In this paper, wepresent our tool CATE (CAusality Tree Extractor), which is able to parse thecomposition of a causal relation as a tree structure .…

## On the Modulus in Matching Vector Codes

A $k$-query locally decodable code (LDC) allows one to encode any$n$-symbol message $x$ as a codeword $C(x) of $N$ symbols . A modulus $m=p_1$ may result in an MVC with $k\leq 2^r$ and $N=\exp(O(O) (1-1/r)) The $m$is {\em good if it is possible to have $k<2^r$. The good numbers yield moreefficient MVCs. Prior to this work, there are only {\em finitely many goodnumbers. All of them were obtained via computer search and have the form$m= p_1p_2$. All of …

## Guided Generation of Cause and Effect

We present a conditional text generation framework that posits sententialexpressions of possible causes and effects . This framework depends on a very large-scale collection of English sentences expressing causal patterns CausalBank . We extend prior work in lexically-constrained decoding to support disjunctive positive constraints .…

## Multi Agent Belief Sharing through Autonomous Hierarchical Multi Level Clustering

Coordination in multi-agent systems is challenging for agile robots such as UAVs, where relative agent positions frequentlychange due to unconstrained movement . This work proposes autonomous hierarchical multi-level clustering (MLC), which forms aclustering hierarchy utilizing decentralized methods . Using observation aggregation, compression, and dissemination, agentsshare local observations throughout the hierarchy, giving every agent a totalsystem belief with spatially dependent resolution and freshness .…

## JEFL Joint Embedding of Formal Proof Libraries

In this paper, we compare apreviously proposed algorithm for matching concepts across libraries with ourunsupervised embedding approach that can help us retrieve similar concepts . Ourapproach is based on the fasttext implementation of Word2Vec, on top of which atree traversal module is added to adapt its algorithm to the representationformat of our data export pipeline .…

## On function homophily of microbial Protein Protein Interaction Networks

We present a new method for assessing homophily in networks whose verticeshave categorical attributes, namely when the vertices of networks comepartitioned into classes . We apply this method to Protein- Protein Interactionnetworks, where vertices correspond to proteins, partitioned according to theyfunctional role, and edges represent potential interactions between proteins .…

## Dynamic RF Combining for Multi Antenna Ambient Energy Harvesting

Ambient radio frequency (RF) energy harvesting (EH) technology is key to realizing self-sustainable, always-on, low-power, massive Internet of Thingsnetworks . We introduce a dynamic RF combining architecture for ambientRF EH use cases . Among the proposed mechanisms, brute force (BF) demands the highest power consumption, while CB requires the highest-resolution phase shifters .…

## Optimal Rates for Nonparametric Density Estimation under Communication Constraints

We consider density estimation for Besov spaces when each sample is quantized to only a limited number of bits . We provide a noninteractive adaptiveestimator that exploits the sparsity of wavelet bases . We show that our estimator is nearly rate-optimal by derivingminimax lower bounds that hold even when interactive protocols are allowed .…

## Improved Text Classification via Contrastive Adversarial Training

We propose a simple and general method to regularize the fine-tuning of Transformer-based encoders for text classification tasks . We generate adversarial examples by perturbing the word embedding of the model and perform contrastive learning on clean and adversary examples .…

## Answer Set Programs for Reasoning about Counterfactual Interventions and Responsibility Scores for Classification

We describe how answer-set programs can be used to declaratively specifycounterfactual interventions on entities under classification, and reason about them . In particular, they can define and compute responsibilityscores as attribution-based explanations for outcomes from classificationmodels . The approach allows for the inclusion of domain knowledge and supports query answering .…

## Accuracy analysis of Educational Data Mining using Feature Selection Algorithm

EducationalData Mining (EDM) helps to measure the accuracy of data using relevant attributes and machine learning algorithms performed . EDM removes irrelevant features without changing the original data . The data set used in this study was taken from Kaggle.com. The results were compared on the basis ofrecall, precision and f-measure to check accuracy of the student data.…

## On Reconfigurability of Target Sets

We study the problem of deciding reconfigurability of target sets of a graph . We prove that Target Set Reconfiguration isPSPACE-complete on bipartite planar graphs of degree $3$ or $4$ and of threshold $2$ and splitgraphs . We present a polynomial-time algorithm for the algorithm on graphsof maximum degree $2 and trees .…

## A Statistical Model of Word Rank Evolution

Thesesignatures suggest unigram frequencies in all languages have changed in amanner inconsistent with a purely neutral evolutionary process . Most of the stopwords andSwadesh words are observed to be stable in ranks across eight languages . High-ranked words tend to be more stable while low-rankedwords tend to have more volatile rates .…

## Into Summarization Techniques for IoT Data Discovery Routing

In this paper, we consider the IoT data discovery problem in very large and growing scale networks . We investigate in depth the routing tablesummarization techniques to support effective and space-efficient IoT datadiscovery routing . Novel summarization algorithms, including alphabeticalbased, hash based, and meaning based summarization, are proposed .…

## Track based Offline Policy Learning for Overtaking Maneuvers with Autonomous Racecars

The rising popularity of driver-less cars has led to the research and development in the field of autonomous racing . In this paper we present the evaluation of a track based offlinepolicy learning approach for autonomous racing. We define specific trackportions and conduct offline experiments to evaluate the probability of anovertaking maneuver based on speed and position of the ego vehicle.…

## Training Electric Vehicle Charging Controllers with Imitation Learning

The problem of coordinating the charging of electric vehicles gains more importance as the number of such vehicles grows . In order to train the controllers, we use the idea of imitation learning . The method is evaluated on realistic data and shows improved performance and training speed compared to similar controllers trained using evolutionary algorithms .…

## A plane wave method based on approximate wave directions for two dimensional Helmholtz equations with large wave numbers

In this paper we present and analyse a high accuracy method for computing wave directions defined in the geometrical optics ansatz of Helmholtz equation . Then we define an “adaptive” plane wave space with small dimensions, in which each plane wave basis function is determined by such an approximate wave direction .…

## Soft Layer Selection with Meta Learning for Zero Shot Cross Lingual Transfer

Multilingual pre-trained contextual embedding models (Devlin et al., 2019) have achieved impressive performance on zero-shot cross-lingual transfer tasks . We propose a novel meta-optimizer tosoft-select which layers of the pre-training model to freeze during fine-tuning . We train the meta-Optimizer by simulating the zero-shoot transfer scenario .…

## On Fair and Efficient Allocations of Indivisible Public Goods

We study fair allocation of indivisible public goods subject to cardinality(budget) constraints . In this model, we have n agents and m available publicgoods, and we want to select $k \leq m$ goods in a fair and efficient manner . We prove that MNW allocations provide fairness guarantees of Proportionality upto one good (Prop1), $1/n$ approximation to Round Robin Share (RRS) and theefficiency guarantee of Pareto Optimality (PO) Further, we show that theproblems of finding MNW or leximin-optimal allocations are NP-hard, even in thecase of constantly many agents, or binary valuations .…

## Debiasing Multilingual Word Embeddings A Case Study of Three Indian Languages

In this paper, we advance the current state-of-the-art method for debiasingmonolingual word embeddings so as to generalize well in a multilingual setting . We believe that our work will open up new opportunitiesin building unbiased downstream NLP applications that are inherently dependent on the quality of the word embeddeddings used .…

## Finite memory strategies in two player infinite games

We study infinite two-player win/lose games where $A,B,W is finite . At each round Player 1 and Player 2 choose one action in $A$ and $B$ respectively . If $W$ is in $Delta^0_2, if the inclusion relation induces a well partial order on the$W_h$’s, if Player 1 has a winning strategy, then she has a finite-memory winning strategy .…

## An artificial intelligence natural language processing pipeline for information extraction in neuroradiology

The use of electronic health records in medical research is difficult because of the unstructured format . We present a natural language processing pipeline for informationextraction of radiological reports in neurology . We train and evaluate a custom languagemodel on a corpus of 150000 reports from National Hospital forNeurology and Neurosurgery, London MRI imaging .…

## Combinatorial Gap Theorem and Reductions between Promise CSPs

A value of a CSP instance is typically defined as a fraction of constraintsthat can be simultaneously met . We show that, for purely combinatorial reasons, a value of anunsolvable instance is bounded away from one . We also show that the gap theorem implies NP-hardness of a gap version of theLayered Label Cover Problem .…

## Efficient Top k Ego Betweenness Search

Betweenness centrality has been recognized as a key indicator for the importance of a vertice in a network . The betweenness of a vertex is hard to compute because it needs to explore all the shortest paths between the other vertices .…

## Peer Selection with Noisy Assessments

In this paper we extend PeerNomination, the most accurate peer reviewing algorithm to date, into WeightedPeerNomination . We show analytically that a weighting scheme can improve the overall accuracy of the selection significantly . We explicitly formulate assessors’ reliability weights in a way that doesn’t violate strategyproofness, and use this information to reweight their scores .…

## Incentivizing Compliance with Algorithmic Instruments

A game-theoretic model to study compliance as dynamic behavior that may change over time . The planner provides each agent with a randomized recommendation that may alter their beliefs and their action selection . Even though the initial agents may be completelynon-compliant, our mechanism can incentivize compliance over time, therebyenabling the estimation of the treatment effect of each treatment, and minimizing the cumulative regret of the planner whose goal is to identify the optimal treatment .…

## Learning Theorem Proving Components

Saturation-style automated theorem provers (ATPs) based on the given clauseprocedure are today the strongest general reasoners for classical first-orderlogic . The clause selection heuristics in such systems are often evaluating clauses in isolation, ignoring other clauses . This has changed recently by equipping the E/ENIGMA system with a graph neural network (GNN) that chooses the next given clause based on its evaluation in the context of previously selected clauses .…

## SkyCell A Space Pruning Based Parallel Skyline Algorithm

Skyline computation is an essential database operation that has many applications in multi-criteria decision making scenarios such as recommendersystems . Existing algorithms have focused on checking point domination, which lack efficiency over large datasets . We propose a grid-based structure that enables grid cell domination checks .…

## CausalBERT Injecting Causal Knowledge Into Pre trained Models with Minimal Supervision

CausalBERT collects the largest scale of causal resource using precise causal patterns and causal embedding techniques . It outperforms all pre-trained models-based state-of-the-art methods, achieving a new causal inference benchmark . In addition, we adopta regularization-based method to preserve the already learned knowledge with an extra regularization term while injecting causal knowledge .…

## Provenance Anonymisation and Data Environments a Unifying Construction

The Anonymisation Decision-making Framework (ADF) operationalizes the riskmanagement of data exchange between organizations . The second edition of ADF has increased its emphasis on modeling data flows . We show how data environments can be implemented within the W3CPROV in four different ways .…

## Using Adversarial Debiasing to Remove Bias from Word Embeddings

Word Embeddings have been shown to contain societal biases present in the original corpora . The method of Adversarial Debiasing was assumed to be similarly superficial, but this is was not verified in previousworks . Using the experiments that demonstrated the shallow removal in othermethods, I show results that suggest Adversary Debiases is more effective at removing bias and thus motivate further investigation on the utility of the method .…

## Answer Set Programs for Reasoning about Counterfactual Interventions and Responsibility Scores for Classification

We describe how answer-set programs can be used to declaratively specifycounterfactual interventions on entities under classification, and reason about them . In particular, they can define and compute responsibilityscores as attribution-based explanations for outcomes from classificationmodels . The approach allows for the inclusion of domain knowledge and supports query answering .…

## A linear Galerkin numerical method for a strongly nonlinear subdiffusion equation

We couple the L1 discretization for Caputo derivative in time with spectralGalerkin method in space to devise a scheme that solves strongly nonlinearsubdiffusion equations . Both the diffusivity and the source are allowed to benonlinear functions of the solution . We prove method’s stability andconvergence with order $2-\alpha$ in time and spectral accuracy in space .…

## Machine learning for assessing quality of service in the hospitality sector based on customer reviews

Paper proposes framework for assessment of the quality of service in the hospitality sector based on theexploitation of customer reviews through natural language processing and machine learning methods . Hotel reviews from Bogot\’a and Madrid are automatically scrapped from Booking.com .…

## Fine Grained Causality Extraction From Natural Language Requirements Using Recursive Neural Tensor Networks

Causal relations (e.g., If A, then B) are prevalent in functional requirements . For various applications of AI4RE, automatically extracting such causal statements are a basic necessity . We lack an approach that is able to extract causal relations from natural language requirements in fine-grainedform .…

## Pushing the Limits Resilience Testing for Mission Critical Machine Type Communication

Interdisciplinary application fields show an increasing need for reliable andresilient wireless communication even under high load conditions . We propose STING, a spatially distributed traffic and interference generation framework . STING is evaluated in a remote control test case of an Unmanned Ground Vehicle that serves as a scout in Search and Rescue missions .…

## Extension of additive valuations to general valuations on the existence of EFX

Envy-freeness is one of the most widely studied notions in fair division . Since envy-free allocations do not always exist when items are indivisible . We show that an EFXallocation always exists (i) when all agents have one of two general valuationsor (ii) when the number of items is at most $n+3$ .…

## Communication Lower Bounds for Nested Bilinear Algorithms

We develop lower bounds on communication in the memory hierarchy or betweenprocessors for nested bilinear algorithms . We build on a previous framework that establishescommunication lower bounds by use of the rank expansion of a matrix . We apply the rankexpansion lower bounds to obtain novel communication lower bounds for nestedToom-Cook convolution, Strassen’s algorithm, and fast algorithms for partiallysymmetric contractions .…

## Provenance Anonymisation and Data Environments a Unifying Construction

The Anonymisation Decision-making Framework (ADF) operationalizes the riskmanagement of data exchange between organizations . The second edition of ADF has increased its emphasis on modeling data flows . We show how data environments can be implemented within the W3CPROV in four different ways .…

## Distribution of Classification Margins Are All Data Equal

Recent theoretical results show that gradient descent on deep neural networks locally maximizes classification margin . This property of the solution however does not fully characterizethe generalization performance . We motivate theoretically and show empiricallythat the area under the curve of the margin distribution on the training set is a good measure of generalization .…

## Comparison of Czech Transformers on Text Classification Tasks

The need for such models emerged from our effort to employ Transformers in our language-specific tasks, but we found the performance of multilingual models to be very limited . Since themultilingual models are usually pre-trained from 100+ languages, most oflow-resourced languages (including Czech) are under-represented in these models .…

## Hilbert Schmidt regularity of symmetric integral operators on bounded domains with applications to SPDE approximations

Regularity estimates for an integral operator with a symmetric continuous kernel on a convex bounded domain are derived . The covariance of a mean-squarecontinuous random field on the domain is an example of such an operator . Theestimates are of the form of Hilbert–Schmidt norms of the integral operatorand its square root, composed with fractional powers of an elliptic operatorequipped with homogeneous boundary conditions of either Dirichlet or Neumanntype .…

## Formal method of synthesis of optimal topologies of computing systems based on projective description of graphs

A deterministic method for synthesizing the interconnect topologies optimized for the required properties is proposed . The method is based on the originaldescription of graphs by projections, on establishing the bijectivecorrespondence of the required . properties and the projection properties of the initial graph, on postulating the corresponding restrictions of modified .…

## THz Transmission meets Untrusted UAV Relaying Trajectory and Communication Co design for Secrecy Energy Efficiency Maximization

Unmanned aerial vehicles (UAVs) and Terahertz (THz) technology are envisioned to play paramount roles in next-generation wireless communications . We present a secure two-phase transmission strategy with cooperative jamming . We assume that the UAV-mounted relay may act, besides providing services, as a potential adversary called the untrusted UAV relay.…