Navigation auf uzh.ch
Navigation auf uzh.ch
For almost 20 years, Google has been at the forefront of building intelligent large scale systems used by millions of users. Within Gmail and Inbox by Gmail apps, machine learning has prominently helped users to reduce spam, to organize their inbox, and to respond to messages quickly.
This talk will provide a high-level overview of how two intelligent Gmail features - Smart Labels and Smart Reply - were developed. Their development will be discussed as a progression through their lifecycle. In the final part of the talk, I will reflect on the commonalities in their development.
Ivo Krka is a senior software engineer at Google. He works on the Gmail Intelligence team on features such as Smart Labels and Smart Reply. Prior to Google, he was a visiting computer scientist at Quandary Peak Research and Infosys.
Krka earned his PhD at the University of Southern California under the supervision of Professor Nenad Medvidovic. His main research interests include requirements engineering, software architecture, and applied machine intelligence.
06.04.2017 - On an Interdisciplinary Empirical Study of Communication, Team Spirit, and Project Success
Speaker: Prof. Dr. Kurt Schneider
Host: Prof. Dr. Martin Glinz
Communication is essential for the success of a software project: Requirements need to be elicited, documented, and design decisions must be spread within a software development team. However, there are symptoms indicating weakening communication, inappropriate communication media, or unsatisfactory information flow. The TeamFLOW project is an interdisciplinary endeavor among software engineers and psychologists, investigating relationships between team spirit, communication behavior, and several aspects of success. By understanding those relationships better, we want to enable project leaders and teams to anticipate problems earlier, by observing key indicators. Prediction and warnings based on communication and team spirit can help keeping a project on track.
In this talk, the main assumptions and hypotheses of the TeamFLOW project are described; key findings are presented, and the research methodology is discussed. TeamFLOW unfolds in two phases: During the current phase, a rather large set of homogenous and comparable student projects is used. Due to the similarity of most variables, the impact of varying communication structures and behaviors can be recognized better. Research methods of software engineers and psychologists needed to be combined and adjusted to the specific situation. Some of the findings will be presented and discussed, such as the role of more or less indirect communication; for example, the concept of “FLOW distance” captures the idea that some teams seem to drift apart, with detrimental consequences for their projects. In the second phase, finding from the homogenous set of projects will be extended towards bigger, more distributed or agile projects with different communication structures.
Kurt Schneider is a full professor of software engineering at Leibniz Universität Hannover, Germany. He studied computer science in Erlangen, received his Doctoral degree from the university of Stuttgart, and held a Postdoctoral position at the interdisciplinary Center for LifeLong Learning and Design at the University of Colorado at Boulder, USA. From 1996 to 2003, he was a researcher and manager at the Daimler Research Center in Ulm, Germany. He conducted research projects with several business units and was the Daimler lead of the international Software Experience Center (SEC) consortium of companies. Since his appointment as professor in 2003, Kurt Schneider´s research interests focus on requirements engineering, effective information flow and communication in software projects, as well as agile and hybrid approaches to software development.
20.04.2017 - Text Analytics to Support Sense-making in Social Media: A Language-Action Perspective
Speaker: Prof. Yilu Zhou, Ph.D.
Host: Prof. Daning Hu, Ph.D.
Social media and online communities provide organizations with new opportunities to support their business-related functions. Despite their various benefits, social media technologies present two important challenges for sense-making. First, online discourse is plagued by incoherent, intertwined conversations that are often difficult to comprehend. Moreover, organizations are increasingly interested in understanding social media participants’ actions and intentions; however, existing text analytics tools mostly focus on the semantic dimension of language. The Language-Action Perspective (LAP) emphasizes pragmatics; not what people say, but rather, what they do with language. Adopting the design science paradigm, we propose a LAP-based text analytics framework to support sense-making in online discourse. The proposed framework is specifically intended to address the two aforementioned challenges associated with sense-making in online discourse: the need for greater coherence and better understanding of actions. We rigorously evaluate a system developed based on the framework in a series of experiments on a test bed encompassing social media data from multiple channels and industries. The results demonstrate the utility of each individual component of the system, and its underlying framework, in comparison with existing benchmark methods. Furthermore, the results of a user experiment involving hundreds of practitioners, and a four-month field experiment in a large organization, underscore the enhanced sense-making capabilities afforded by text analytics grounded in LAP principles. The results have important implications for online sense-making and social media analytics.
Yilu Zhou, PhD, is an associate professor at the Gabelli School of Business. Her research interests include business intelligence, web/text/data mining, multilingual knowledge discovery and human-computer interaction. Most specifically, she investigates and explores computational, intelligent and automatic ways to discover interesting and useful patterns in news articles, web sites, forums and other social media. Before joining Fordham University, Dr. Zhou was an assistant professor at George Washington University. She received a PhD in management information systems at the University of Arizona, where she also was a research associate at the Artificial Intelligence Lab. She received her BS in computer science from Shanghai Jiaotong University. Dr. Zhou has published work in academic journals including the Journal of the American Society for Information Science and Technology, IEEE Intelligent Systems and Decision Support Systems. She is a recipient of a NSF grant to study mobile app maturity rating for children. She has taught various courses related to business intelligence, data management and business programming.
27.04.2017 - Finding the Thread (Andean Style)
Speaker: Prof. Dr. Sven Helmer
Host: Prof. Dr. Michael Böhlen
Andean textiles are products of one of the richest, oldest and continuous weaving traditions in the world. Understanding the knowledge and practice of textile production as a form of cultural heritage is particularly relevant in the Andean context due to erosion of clothing traditions, reuse of traditional textiles on commodities targeted at the tourism market, and loss of knowledge embedded in textile production. We modeled the knowledge of domain experts and represented information about the textile objects in the form of an OWL ontology. We also developed a suite of search facilities, making the digitized fabrics accessible to domain experts and researchers in the area.
Sven Helmer is an Associate Professor at the Faculty of Computer Science at the Free University of Bozen-Bolzano, Italy, after holding a position as Senior Lecturer at Birkbeck, University of London. He obtained a PhD from the University of Mannheim, Germany, an MSc in Computer Science from the University of Karlsruhe, Germany, and also spent some time as a visiting professor at the University of Heidelberg. He is currently teaching courses on databases and information security; his research interests include database systems, cloud computing, Raspberry Pis, query optimization, route planning, as well as interdisciplinary research in the areas of information systems and ethnography. He has published more than 70 peer-reviewed papers and book chapters.
04.05.2017 - Positive Impact Games
Speaker: Prof. Dr. Peter Purgathofer
Host: Prof. Dr. Renato Pajarola
The term "positive impact games" evokes many associations. It also implicates a seductive vision of games that are administered like medication to repair or mitigate an individual - or a society's - pain, damage or shortcomings. The talk includes a critical reflection on a number of projects we have worked on in recent years within the "positive impact games lab" at the Institute of Design and Assessment of Technology, TU Wien, as well as an attempt to indentify current scientific questions and perspectives in the field.
Prof. Dr. Peter Purgathofer majored in informatics in Vienna, and has since focused on the issue of design within informatics. In the course of his work he founded and led the UID-Lab, a cooperation with GP designpartners, where they worked on numerous national and international design projects between 1996 and 2004. Among the clients were Ars Electronica center, the CD-Labor für Software Research and the German Sparkassen-Informatikzentrum. UID-Lab has spun off into a design enterprise and is now part of GP designpartners. Prof. Purgathofer is a (co-)winner of an award of distinction at the Prix Ars Electronica 2000 in the category .net, together with Rich Bergerand, Volker Christian (and many others) for the project telezone. Since early 2005 he has been an associate professor for interactive systems, and his habilitation discusses history, theory and practice of design methodology, focused on the design of interactive software. Since 2007, he has been the coordinator for the Media Informatics bachelor and master programs at the Vienna University of Technology. In 2012, he co-founded Positive Impact Games Lab (pig·lab) together with Fares Kayali.
18.05.2017 - An Open Access Wireless Market
Speaker: Prof. Peter Cramton, Ph.D.
Host: Prof. Sven Seuken, Ph.D.
We present a market design for an open access wireless market. Open access means that in real time, network capacity cannot be withheld—network throughput is priced dynamically by the marginal demand during congestion. In unconstrained times and locations, a nominal fee is paid for network throughput. As in electricity markets, the real-time market provides the foundation for forward markets. Monthly forwards are auctioned before the start of each month; yearly forwards are auctioned before the start of each year. Market participants, both operators and traders, take positions in forward auctions to manage risk and optimize portfolios. Deviations from forward positions are settled at real-time prices based on actual use. The independent system operator runs the network and conducts the real-time, monthly, and yearly auctions of network throughput. An independent market monitor observes the market, identifies problems, and suggests solutions. A board—including affiliated directors representing important stakeholders together with independent directors with subject matter expertise—governs the market. A goal of the market is to provide a secure, robust, wide-coverage platform for mobile communications supporting public safety and universal service. Public safety has pre-emptive rights during emergencies and otherwise has economic use like wholesale operators. A complementary goal is competition. The open access provision brings vibrant competition through low-cost, non-discriminatory entry into the wireless market. The market provides a natural remedy for mergers, allowing operational efficiency gains while increasing competition. Critical funding is provided through efficient congestion pricing that balances supply and demand at every time and location. The market, enabled by flexible handsets and the LTE technology, radically reforms current spectrum policy. The
market coexists and complements the dedicated networks of incumbent carriers, promoting efficient spectrum use and essential innovation.
Prof. Peter Cramton, Ph.D., is Professor of Economics at the University of Maryland and European University Institute, and on the International Faculty at the University of Cologne. Since 1983, he has conducted research on auction theory and practice. This research appears in the leading economics journals. The main focus is the design of auctions for many related items. Applications include spectrum, energy, and financial auctions. On the practical side, he is an independent director on the board of the Electric Reliability Council of Texas, chief economist of Rivada, and chairman of Market Design Inc., an economics consultancy founded in 1995, focusing on the design of auction and matching markets. Since 1993, he has advised 12 governments and 40 bidders in spectrum auctions. He is a co-inventor of the spectrum auction design used in Canada, Australia, and many European countries to auction 4G spectrum. Since 2001, he has played a lead role in the design and implementation of electricity and gas auctions in North America, South America, and Europe. He has advised on the design of carbon auctions in Europe, Australia, and the United Sates, including conducting the world’s first greenhouse-gas auction held in the UK in 2002. He has led the development of innovative auctions in new applications, such as auctions for airport slots, wind rights, diamonds, medical equipment, and Internet top-level domains. He received his B.S. in Engineering from Cornell University in 1980 and his Ph.D. in Business from Stanford University in 1984.
