Abstract The standard approach to evaluate text anonymization methods consists of comparing their outcomes with the anonymization performed by human experts. The degree of privacy protection attained is then measured with the IR-based recall metric, which expresses the proportion of re-identifying terms that were correctly detected by the anonymization method. However, the use of recall to estimate the degree of privacy protection suffers from several limitations. The first is that it assigns a uniform weight to each re-identifying term, thereby ignoring the fact that some missed re-identifying terms may have a larger influence on the disclosure risk than others. Furthermore, IR-based metrics assume the existence of a single gold standard annotation. This assumption does not hold for text anonymization, where several maskings (each one encompassing a different combination of terms) could be equally valid to prevent disclosure. Finally, those metrics rely on manually anonymized datasets, which are inherently subjective and may be prone to various errors, omissions and inconsistencies. To tackle these issues, we propose an automatic re-identification attack for (anonymized) texts that provides a realistic assessment of disclosure risks. Our method follows a similar premise as the well-known record linkage methods employed to evaluate anonymized structured data, and leverages state-of-the-art deep learning language models to exploit the background knowledge available to potential attackers. We also report empirical evaluations of several well-known methods and tools for text anonymization. Results show significant re-identification risks for all methods, including also manual anonymization efforts.KeywordsText anonymizationRe-identification riskLanguage modelsBERT

Ever-increasing data size raises many challenges for scientific data analysis. Particularly in cosmological N-body simulation, finding the center of a dark matter halo suffers heavily from the large computational cost associated with the large number of particles (up to 20 million). In this work, we exploit the latent structure embed in a halo, and we propose a hierarchical approach to approximate the exact gravitational potential calculation for each particle in order to more efficiently find the halo center. Tests of our method on data from N-body simulations show that in many cases the hierarchical algorithm performs significantly faster than existing methods with a desirable accuracy.KeywordsN-body simulationHierarchical analysisClusteringHalo center

This paper introduces the motivation and aim of the 1st International Symposium on Modelling and Knowledge Management for Sustainable Development MoKMaSD 2012, inspired by the POST-2015 UN Development Agenda. Then the keynote paper and the four contributed papers presented at the Symposium are summarised and related to the POST-2015 UN Development Agenda.

Spiking Neural Network (SNN) is the most recent computational model that can emulate the behaviors of biological neuron system. This paper highlights and discusses an efficient hardware architecture for the hardware SNNs, which includes a layer-level tile architecture (LTA) for the neurons and synapses, and a novel routing architecture (NRA) for the interconnections between the neuron nodes. In addition, a visualization performance monitoring platform is designed, which is used as functional verification and performance monitoring for the SNN hardware system. Experimental results demonstrate that the proposed architecture is feasible and capable of scaling to large hardware multilayer SNNs.

In machine learning, the accuracy of the system depends upon classification result. Classification accuracy plays an imperative role in various domains. Non-parametric classifier like K-Nearest Neighbor (KNN) is the most widely used classifier for pattern analysis. Besides its easiness, simplicity and effectiveness characteristics, the main problem associated with KNN classifier is the selection of number of nearest neighbors i.e. ‘k’ for computation. At present it is hard to find the optimal value of ‘k’ using any statistical algorithm, which gives perfect accuracy in terms of low misclassification error rate.

This paper presents a combination of several automated reasoning and proof presentation tools with the Mizar system for formalization of mathematics. The combination forms an online service called MizAR, similar to the SystemOnTPTP service for first-order automated reasoning. The main differences to SystemOnTPTP are the use of the Mizar language that is oriented towards human mathematicians (rather than the pure first-order logic used in SystemOnTPTP), and setting the service in the context of the large Mizar Mathematical Library of previous theorems, definitions, and proofs (rather than the isolated problems that are solved in SystemOnTPTP). These differences poses new challenges and new opportunities for automated reasoning and for proof presentation tools. This paper describes the overall structure of MizAR, and presents the automated reasoning systems and proof presentation tools that are combined to make MizAR a useful mathematical service.

Objective: To automatically detect duplicate citations in a bibliographical database. Background: Citations retrieved from multiple search databases have different forms making manual and automatic detection of duplicates difficult. Existing methods rely on fuzzy-similarity measures which are error-prone. Methods: We analysed four pairs of original search results from MEDLINE and EMBASE that were used to create systematic reviews. An automatic tool deduplicated citations by first enriching citations with Digital Object Identifiers DOI, and/or other unique identifiers. Duplication of records was then determined by comparing these unique identifiers. We compared our method with the duplicate detection function of a popular citation management desktop application in several configurations. Results: Citation Enrichment identified 93i¾?% range 86i¾?%---100i¾?% of the duplicates indexed online and erroneously marked 3i¾?% range 0i¾?%---6i¾?% documents as duplicates. The citation management application found 68i¾?% range 64i¾?%---72i¾?% without error using default setting. When set for highest deduplication, the citation management application found 94i¾?% of duplicates range 77i¾?%---100i¾?% and 4i¾?% error range 0i¾?%---8i¾?%. Conclusion: Citation enrichment using unique identifiers enhances automatic deduplication. On its own, the approach seems slightly superior to tools that compare citations without enrichment. Methods that combine citation enrichment with existing fuzzy-matching may substantially reduce resource requirements of evidence synthesis.

One of the most popular forms of learning is through reading and for years we have used hard copy documents as the main material to learn. With the advent of the Internet and the fast development of new technologies, new tools have been developed to assist the learning process. However, reading is still the main learning method that is an individual activity. In this paper we propose a highlighting tool that enables the reading and learning process to become a collaborative and shared activity. In other words, the highlighting tool supports the so-called active-reading, a well-known and efficient means of learning. The highlighting tool brings to the digital environment the same metaphor of the traditional highlight marker and puts it in a social context. It enables users to emphasize certain portions of digital learning objects. Furthermore, it provides students, tutors, course coordinators and educational institutions new possibilities in the teaching and learning process. In this work we expose the first quantitative and qualitative results regarding the use of the highlight tool by over 750 students through 8 weeks of courses.

Differentiation in human beings is the act of perceiving the difference in or between objects. In other words, it is the mental process taking place to discriminate one thing from others, a common task performed by a person on a very regular basis. Making such differentiations, small or large, easy or hard, still requires a combination of cognitive processes to occur across various parts of the human brain. In this paper, an EEG-based BCI experiment was organized to study the detection of such cognitive processes. Utilizing a machine learning tool, Artificial Neural Networks, to aid in analyzing the acquired dataset, a high correct classification rate was achieved, confirming that it is possible to computationally detect these differentiation activities from EEG signals.