1,722,120 research outputs found
EFSG: Evolutionary Fooling Sentences Generator
Full text linkLarge pre-trained language representation models (LMs) have recently collected a huge number of successes in many NLP tasks. In 2018 BERT, and later its successors (e.g. RoBERTa), obtained state-of-the-art results in classical benchmark tasks, such as GLUE. Works about adversarial attacks have been published to test their generalization proprieties and robustness. In this study, we propose Evolutionary Fooling Sentences Generator (EFSG), a black-box task-agnostic adversarial attack algorithm designed in an evolutionary fashion to generate false-positive sentences for binary classification tasks. We successfully apply EFSG to single-sentence (CoLA) and sentence-pair (MRPC) classification tasks, on BERT and RoBERTa. Results prove the presence of weak spots in state-of-the-art LMs. To complete the analysis, we perform transferability tests and ablation study. Finally, adversarial training helps as a data augmentation defence approach against EFSG, obtaining stronger improved models with no loss of accuracy
Sensor-Assisted Cooperative Localization and Communication in Multi-agent Networks
No full textThis brief highlights research advances on cooperative techniques for localization and communication. These two macro trends are investigated in the general context of mobile multi-agent networks for situational awareness applications, where time-varying agents of unknown locations are asked to fulfill positioning and information sharing tasks. Cooperative localization is conceived for both active and passive agents, i.e., targets to be detected and localized, and it is analyzed in vehicular and maritime environments. Communication is investigated for vehicular scenarios, where vehicles are requested to share massive data in the perspective development of connected and automated mobility systems. Both research areas rely on the integration of heterogeneous sensors and communication. Specifically, it is studied how to improve localization by exploring communication techniques as well as how to enhance communication performances by extracting information from perception sensors. The dynamic environment of multi-agent systems calls for robust, flexible and adaptive techniques, capable of profitably fuse different types of information, and the outcomes of these researches show how a statistical approach based on cooperation guarantees higher resilience, reliability and confidence
Brand community analysis on social networks using graph representation learning
No full textIn a world more and more connected, new and complex interaction patterns can be extracted in the communication between people. This is extremely valuable for brands that can better understand the interests of users and the trends on social media to better target their products. In this paper, we aim to analyze the communities that arise around commercial brands on social networks to understand the meaning of similarity, collaboration, and interaction among users. We exploit the network that builds around the brands by encoding it into a graph model. We build a social network graph, considering user nodes and friendship relations; then we compare it with a heterogeneous graph model, where also posts and hashtags are considered as nodes and connected to the different node types; we finally build also a reduced network, generated by inducing direct user-to-user connections through the intermediate nodes (posts and hashtags). These different variants are encoded using graph representation learning, which generates a numerical vector for each node. Machine learning techniques are applied to these vectors to extract valuable insights for each user and for the communities they belong to. In the paper, we report on our experiments performed on an emerging fashion brand on Instagram, and we show that our approach is able to discriminate potential customers for the brand, and to highlight meaningful sub-communities composed by users that share the same kind of content on social networks
Wide-spectrum characterization of long-running political phenomena on social media: The brexit case
No full textIn this study, we propose a wide-spectrum analysis of long-running political events on social media, with reference to an interesting real-world international case: the so-called Brexit, the process through which the United Kingdom activated the option of leaving the European Union. In this study, we model the users participating in 33 months of Twitter debate, covering their behaviour and demographics. By using publicly shared tweets, we developed a stance classification model to evaluate the change of stance over time. We also extracted the key topics of the long-running debate, studying which political side have discussed them most and what is the general sentiment on each. We also revealed the participation of bot accounts, and we found that the higher the bot score, the more likely the account is in a pro-Leave position. We conclude our study with a temporal and comparative analysis of politicians' social media accounts
Voice-Based Virtual Assistants for User Interaction Modeling
No full textIn this work, we propose a virtual assistant that allows building models by means of voice commands. To demonstrate the generality of the approach, we describe three alternative strategies that apply voice-based support at three levels of detail: a fully-guided strategy; a pattern-based strategy; and an element-based strategy. We describe our implementation experience with the development of a design assistant covering the three strategies described above for OMG’s IFML (Interaction Flow Modeling Language), in the context of user interaction design, including the integration with the Amazon Alexa assistant. We report our results that show how the assistant can bring advantages in terms of productivity
DISPERSION OF THE TWO-PHOTON ABSORPTION COEFFICIENT IN ZnSe
No full textThe dispersion of the degenerate two-photon absorption coefficient is measured in ZnSe in the energy region from 2.66 to 3.52 eV, corresponding to 0.5 < hv/E-g < 0.66. Our results clearly show the sizeable contribution of the split-off band to the strength of the nonlinear absorption. By contrasting our measurement with several theoretical calculations, we find that at least the lowest four-bands should be included in the band-structure model of the two-photon absorption. General scaling laws for the calculation of the two-photon absorption coefficient provide a good estimate of the strength of the absorption but fail to account for details of the spectrum
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
"Decomposition of bivariate inequality indices by attributes” by Abul Naga and Geoffard, Economics Letters 90 (2006), pp. 362-367
No full textWe correct the generalized version of the decomposition of multivariate inequality indices by attributes
proposed by Abul Naga and Geoffard (Abul Naga, R. H. and Geoffard, P. Y., 2006. Decomposition of bivariate
inequality indices by attributes. Economic Letters 90, pp. 362–367)
- …
