TY - CONF AB - Recommender Agents (RAs) facilitate consumers’ online purchase decisions for complex, multi-attribute products. As not all combinations of attribute levels can be obtained, users are forced into trade-offs. The exposure of trade-offs in a RA has been found to affect consumers’ perceptions. However, little is known about how different preference elicitation methods in RAs affect consumers by varying degrees of trade-off exposure. We propose a research model that investigates how different levels of trade-off exposure cognitively and affectively influence consumers’ satisfaction with RAs. We operationalize these levels in three different RA types and test our hypotheses in a laboratory experiment with 116 participants. Our results indicate that with increasing tradeoff exposure, perceived enjoyment and perceived control follow an inverted Ushaped relationship. Hence, RAs using preference elicitation methods with medium trade-off exposure yield highest consumer satisfaction. This contributes to the understanding of trade-offs in RAs and provides valuable implications to e-commerce practitioners. AU - Schuhbeck, Veronika AU - Siegfried, Nils AU - Dorner, Verena AU - Benlian, Alexander AU - Scholz, Michael AU - Schryen, Guido ID - 6514 KW - Recommender Agents KW - Preference Elicitation Method KW - Trade-off Exposure KW - Customer Satisfaction T2 - Proceedings of the 14. Internationale Tagung Wirtschaftsinformatik TI - Walking the Middle Path: How Medium Trade-off Exposure Leads to Higher Consumer Satisfaction in Recommender Agents ER - TY - CONF AU - Afifi, Haitham AU - Karl, Holger ID - 6860 T2 - 2019 16th IEEE Annual Consumer Communications & Networking Conference (CCNC2019) TI - Power Allocation with a Wireless Multi-cast Aware Routing for Virtual Network Embedding ER - TY - GEN AB - In this work we describe our results achieved in the ProtestNews Lab at CLEF 2019. To tackle the problems of event sentence detection and event extraction we decided to use contextualized string embeddings. The models were trained on a data corpus collected from Indian news sources, but evaluated on data obtained from news sources from other countries as well, such as China. Our models have obtained competitive results and have scored 3rd in the event sentence detection task and 1st in the event extraction task based on average F1-scores for different test datasets. AU - Skitalinskaya, Gabriella AU - Klaff, Jonas AU - Spliethöver, Maximilian ID - 16847 TI - CLEF ProtestNews Lab 2019: Contextualized Word Embeddings for Event Sentence Detection and Event Extraction VL - 2380 ER - TY - CONF AB - We present an unsupervised training approach for a neural network-based mask estimator in an acoustic beamforming application. The network is trained to maximize a likelihood criterion derived from a spatial mixture model of the observations. It is trained from scratch without requiring any parallel data consisting of degraded input and clean training targets. Thus, training can be carried out on real recordings of noisy speech rather than simulated ones. In contrast to previous work on unsupervised training of neural mask estimators, our approach avoids the need for a possibly pre-trained teacher model entirely. We demonstrate the effectiveness of our approach by speech recognition experiments on two different datasets: one mainly deteriorated by noise (CHiME 4) and one by reverberation (REVERB). The results show that the performance of the proposed system is on par with a supervised system using oracle target masks for training and with a system trained using a model-based teacher. AU - Drude, Lukas AU - Heymann, Jahn AU - Haeb-Umbach, Reinhold ID - 11965 T2 - INTERSPEECH 2019, Graz, Austria TI - Unsupervised training of neural mask-based beamforming ER - TY - CONF AB - We propose a training scheme to train neural network-based source separation algorithms from scratch when parallel clean data is unavailable. In particular, we demonstrate that an unsupervised spatial clustering algorithm is sufficient to guide the training of a deep clustering system. We argue that previous work on deep clustering requires strong supervision and elaborate on why this is a limitation. We demonstrate that (a) the single-channel deep clustering system trained according to the proposed scheme alone is able to achieve a similar performance as the multi-channel teacher in terms of word error rates and (b) initializing the spatial clustering approach with the deep clustering result yields a relative word error rate reduction of 26% over the unsupervised teacher. AU - Drude, Lukas AU - Hasenklever, Daniel AU - Haeb-Umbach, Reinhold ID - 12874 T2 - ICASSP 2019, Brighton, UK TI - Unsupervised Training of a Deep Clustering Model for Multichannel Blind Source Separation ER - TY - CONF AB - Signal dereverberation using the Weighted Prediction Error (WPE) method has been proven to be an effective means to raise the accuracy of far-field speech recognition. First proposed as an iterative algorithm, follow-up works have reformulated it as a recursive least squares algorithm and therefore enabled its use in online applications. For this algorithm, the estimation of the power spectral density (PSD) of the anechoic signal plays an important role and strongly influences its performance. Recently, we showed that using a neural network PSD estimator leads to improved performance for online automatic speech recognition. This, however, comes at a price. To train the network, we require parallel data, i.e., utterances simultaneously available in clean and reverberated form. Here we propose to overcome this limitation by training the network jointly with the acoustic model of the speech recognizer. To be specific, the gradients computed from the cross-entropy loss between the target senone sequence and the acoustic model network output is backpropagated through the complex-valued dereverberation filter estimation to the neural network for PSD estimation. Evaluation on two databases demonstrates improved performance for on-line processing scenarios while imposing fewer requirements on the available training data and thus widening the range of applications. AU - Heymann, Jahn AU - Drude, Lukas AU - Haeb-Umbach, Reinhold AU - Kinoshita, Keisuke AU - Nakatani, Tomohiro ID - 12875 T2 - ICASSP 2019, Brighton, UK TI - Joint Optimization of Neural Network-based WPE Dereverberation and Acoustic Model for Robust Online ASR ER - TY - CONF AB - In this paper, we present libDirectional, a MATLAB library for directional statistics and directional estimation. It supports a variety of commonly used distributions on the unit circle, such as the von Mises, wrapped normal, and wrapped Cauchy distributions. Furthermore, various distributions on higher-dimensional manifolds such as the unit hypersphere and the hypertorus are available. Based on these distributions, several recursive filtering algorithms in libDirectional allow estimation on these manifolds. The functionality is implemented in a clear, well-documented, and object-oriented structure that is both easy to use and easy to extend. AU - Kurz, Gerhard AU - Gilitschenski, Igor AU - Pfaff, Florian AU - Drude, Lukas AU - Hanebeck, Uwe D. AU - Haeb-Umbach, Reinhold AU - Siegwart, Roland Y. ID - 12876 T2 - Journal of Statistical Software 89(4) TI - Directional Statistics and Filtering Using libDirectional ER - TY - CONF AB - One of the major challenges in implementing wireless virtualization is the resource discovery. This is particularly important for the embedding-algorithms that are used to distribute the tasks to nodes. MARVELO is a prototype framework for executing different distributed algorithms on the top of a wireless (802.11) ad-hoc network. The aim of MARVELO is to select the nodes for running the algorithms and to define the routing between the nodes. Hence, it also supports monitoring functionalities to collect information about the available resources and to assist in profiling the algorithms. The objective of this demo is to show how MAVRLEO distributes tasks in an ad-hoc network, based on a feedback from our monitoring tool. Additionally, we explain the work-flow, composition and execution of the framework. AU - Afifi, Haitham AU - Karl, Holger AU - Eikenberg, Sebastian AU - Mueller, Arnold AU - Gansel, Lars AU - Makejkin, Alexander AU - Hannemann, Kai AU - Schellenberg, Rafael ID - 12882 KW - WSN KW - virtualization KW - VNE T2 - 2019 IEEE Wireless Communications and Networking Conference (WCNC) (IEEE WCNC 2019) (Demo) TI - A Rapid Prototyping for Wireless Virtual Network Embedding using MARVELO ER - TY - JOUR AB - We formulate a generic framework for blind source separation (BSS), which allows integrating data-driven spectro-temporal methods, such as deep clustering and deep attractor networks, with physically motivated probabilistic spatial methods, such as complex angular central Gaussian mixture models. The integrated model exploits the complementary strengths of the two approaches to BSS: the strong modeling power of neural networks, which, however, is based on supervised learning, and the ease of unsupervised learning of the spatial mixture models whose few parameters can be estimated on as little as a single segment of a real mixture of speech. Experiments are carried out on both artificially mixed speech and true recordings of speech mixtures. The experiments verify that the integrated models consistently outperform the individual components. We further extend the models to cope with noisy, reverberant speech and introduce a cross-domain teacher–student training where the mixture model serves as the teacher to provide training targets for the student neural network. AU - Drude, Lukas AU - Haeb-Umbach, Reinhold ID - 12890 JF - IEEE Journal of Selected Topics in Signal Processing TI - Integration of Neural Networks and Probabilistic Spatial Models for Acoustic Blind Source Separation ER - TY - JOUR AU - Hammer, Manfred AU - Ebers, Lena AU - Förstner, Jens ID - 12908 JF - Journal of the Optical Society of America B KW - tet_topic_waveguides SN - 0740-3224 TI - Oblique quasi-lossless excitation of a thin silicon slab waveguide: a guided-wave variant of an anti-reflection coating VL - 36 ER - TY - CONF AB - More and more management and orchestration approaches for (software) networks are based on machine learning paradigms and solutions. These approaches depend not only on their program code to operate properly, but also require enough input data to train their internal models. However, such training data is barely available for the software networking domain and most presented solutions rely on their own, sometimes not even published, data sets. This makes it hard, or even infeasible, to reproduce and compare many of the existing solutions. As a result, it ultimately slows down the adoption of machine learning approaches in softwarised networks. To this end, we introduce the "softwarised network data zoo" (SNDZoo), an open collection of software networking data sets aiming to streamline and ease machine learning research in the software networking domain. We present a general methodology to collect, archive, and publish those data sets for use by other researches and, as an example, eight initial data sets, focusing on the performance of virtualised network functions. AU - Peuster, Manuel AU - Schneider, Stefan Balthasar AU - Karl, Holger ID - 15371 T2 - IEEE/IFIP 15th International Conference on Network and Service Management (CNSM) TI - The Softwarised Network Data Zoo ER - TY - CONF AB - Offloading packet processing tasks to programmable switches and/or to programmable network interfaces, so called “SmartNICs”, is one of the key concepts to prepare softwarized networks for the high traffic demands of the future. However, implementing network functions that make use of those offload- ing technologies is still challenging and usually requires the availability of specialized hardware. It becomes even harder if heterogeneous services, making use of different offloading and network virtualization technologies, should be developed. In this paper, we introduce FOP4 (Function Offloading Pro- totyping with P4), a novel prototyping platform that allows to prototype heterogeneous software network scenarios, including container-based, P4-switch-based, and SmartNIC-based network functions. The presented work substantially extends our existing Containernet platform with the means to prototype offloading scenarios. Besides presenting the platform’s system design, we evaluate its scalability and show that it can run scenarios with more than 64 P4 switch or SmartNIC nodes on a single laptop. Finally, we presented a case study in which we use the presented platform to prototype an extended in-band network telemetry use case. AU - Moro, Daniele AU - Peuster, Manuel AU - Karl, Holger AU - Capone, Antonio ID - 15373 T2 - IEEE Conference on Network Function Virtualization and Software Defined Networks (NFV-SDN) TI - FOP4: Function Offloading Prototyping in Heterogeneous and Programmable Network Scenarios ER - TY - CONF AB - Emulation platforms supporting Virtual Network Functions (VNFs) allow developers to rapidly prototype network services. None of the available platforms, however, supports experimenting with programmable data planes to enable VNF offloading. In this demonstration, we show FOP4, a flexible platform that provides support for Docker-based VNFs, and VNF offloading, by means of P4-enabled switches. The platform provides interfaces to program the P4 devices and to deploy network functions. We demonstrate FOP4 with two complex example scenarios, demonstrating how developers can exploit data plane programmability to implement network functions. AU - Moro, Daniele AU - Peuster, Manuel AU - Karl, Holger AU - Capone, Antonio ID - 15374 T2 - IEEE Conference on Network Function Virtualization and Software Defined Networks (NFV-SDN) TI - Demonstrating FOP4: A Flexible Platform to Prototype NFV Offloading Scenarios ER - TY - JOUR AU - Jochen Baumeister ID - 15416 JF - Quick And Easy Journal Title TI - New Quick And Easy Publication - Will be edited by LibreCat team ER - TY - CONF AB - Connectionist temporal classification (CTC) is a sequence-level loss that has been successfully applied to train recurrent neural network (RNN) models for automatic speech recognition. However, one major weakness of CTC is the conditional independence assumption that makes it difficult for the model to learn label dependencies. In this paper, we propose stimulated CTC, which uses stimulated learning to help CTC models learn label dependencies implicitly by using an auxiliary RNN to generate the appropriate stimuli. This stimuli comes in the form of an additional stimulation loss term which encourages the model to learn said label dependencies. The auxiliary network is only used during training and the inference model has the same structure as a standard CTC model. The proposed stimulated CTC model achieves about 35% relative character error rate improvements on a synthetic gesture keyboard recognition task and over 30% relative word error rate improvements on the Librispeech automatic speech recognition tasks over a baseline model trained with CTC only. AU - Heymann, Jahn AU - Khe Chai Sim, Bo Li ID - 15812 T2 - ICASSP 2019, Brighton, UK TI - Improving CTC Using Stimulated Learning for Sequence Modeling ER - TY - CONF AB - Despite the strong modeling power of neural network acoustic models, speech enhancement has been shown to deliver additional word error rate improvements if multi-channel data is available. However, there has been a longstanding debate whether enhancement should also be carried out on the ASR training data. In an extensive experimental evaluation on the acoustically very challenging CHiME-5 dinner party data we show that: (i) cleaning up the training data can lead to substantial error rate reductions, and (ii) enhancement in training is advisable as long as enhancement in test is at least as strong as in training. This approach stands in contrast and delivers larger gains than the common strategy reported in the literature to augment the training database with additional artificially degraded speech. Together with an acoustic model topology consisting of initial CNN layers followed by factorized TDNN layers we achieve with 41.6% and 43.2% WER on the DEV and EVAL test sets, respectively, a new single-system state-of-the-art result on the CHiME-5 data. This is a 8% relative improvement compared to the best word error rate published so far for a speech recognizer without system combination. AU - Zorila, Catalin AU - Boeddeker, Christoph AU - Doddipatla, Rama AU - Haeb-Umbach, Reinhold ID - 15816 T2 - ASRU 2019, Sentosa, Singapore TI - An Investigation Into the Effectiveness of Enhancement in ASR Training and Test for Chime-5 Dinner Party Transcription ER - TY - CONF AU - Müller, Jens AU - Brinkmann, Marcus AU - Poddebniak, Damian AU - Böck, Hanno AU - Schinzel, Sebastian AU - Somorovsky, Juraj AU - Schwenk, Jörg ID - 15908 SN - 978-1-939133-06-9 T2 - 28th {USENIX} Security Symposium ({USENIX} Security 19) TI - "Johnny, you are fired!" -- Spoofing OpenPGP and S/MIME Signatures in Emails ER - TY - CONF AU - Merget, Robert AU - Somorovsky, Juraj AU - Aviram, Nimrod AU - Young, Craig AU - Fliegenschmidt, Janis AU - Schwenk, Jörg AU - Shavitt, Yuval ID - 15909 SN - 978-1-939133-06-9 T2 - 28th {USENIX} Security Symposium ({USENIX} Security 19) TI - Scalable Scanning and Automatic Classification of TLS Padding Oracle Vulnerabilities ER - TY - CONF AU - Heindorf, Stefan AU - Scholten, Yan AU - Engels, Gregor AU - Potthast, Martin ID - 14568 T2 - INFORMATIK TI - Debiasing Vandalism Detection Models at Wikidata (Extended Abstract) ER - TY - CONF AB - Multi-talker speech and moving speakers still pose a significant challenge to automatic speech recognition systems. Assuming an enrollment utterance of the target speakeris available, the so-called SpeakerBeam concept has been recently proposed to extract the target speaker from a speech mixture. If multi-channel input is available, spatial properties of the speaker can be exploited to support the source extraction. In this contribution we investigate different approaches to exploit such spatial information. In particular, we are interested in the question, how useful this information is if the target speaker changes his/her position. To this end, we present a SpeakerBeam-based source extraction network that is adapted to work on moving speakers by recursively updating the beamformer coefficients. Experimental results are presented on two data sets, one with articially created room impulse responses, and one with real room impulse responses and noise recorded in a conference room. Interestingly, spatial features turn out to be advantageous even if the speaker position changes. AU - Heitkaemper, Jens AU - Feher, Thomas AU - Freitag, Michael AU - Haeb-Umbach, Reinhold ID - 14822 T2 - International Conference on Statistical Language and Speech Processing 2019, Ljubljana, Slovenia TI - A Study on Online Source Extraction in the Presence of Changing Speaker Positions ER -