Logo des Repositoriums

Auflistung nach Autor:in "Gertz, Michael"

1 - 8 von 8
  • Zeitschriftenartikel
    Der Lehrstuhl für Datenbanksysteme am Institut für Informatik der Universität Heidelberg
    (Datenbank-Spektrum: Vol. 10, No. 3, 2010) Gertz, Michael
  • Zeitschriftenartikel
    Editorial
    (Datenbank-Spektrum: Vol. 12, No. 3, 2012) Gertz, Michael; Müller, Wolfgang
  • Textdokument
    EPINetz: Exploration of Political Information Networks
    (INFORMATIK 2021, 2021) Ziegler, John; Brand, Alexander; Freyberg, Julian; König, Tim; Schünemann, Wolf; Walther, Marina; Gertz, Michael
    Different societal challenges, such as information overload, emerge due to the digital transformation of the media landscape. This also demands new competencies from citizens that often lack the means to contextualize arguments or actors and to understand their interrelationships in complex topics. The EPINetz project is an approach to bridge the outlined skills gap by developing an appropriate political information system. It provides access to political news collected from multiple data sources, including social media, and offers various network exploration capabilities. Different entities such as political actors or topics are extracted from collected data and shown within their respective contexts modelled as weighted and time-varying information networks. Thereby, interested citizens and especially schoolchildren can discover current political topics and understand relationships between relevant entities.
  • Konferenzbeitrag
    EvenPers: event-based person exploration and correlation
    (Datenbanksysteme für Business, Technologie und Web (BTW) 2049, 2013) Kapp, Christian; Strötgen, Jannik; Gertz, Michael
    Searching for people on the Internet is one of the most frequent search activities. In this paper, we present EvenPers, a system for the event-based exploration of persons and person similarities. We address challenges such as cross-document person name normalization and present a novel approach to calculate person similarities based on their event information. In our demonstration, we show several exploration scenarios illustrating the usefulness of EvenPers and its exciting functionality.
  • Konferenzbeitrag
    In-network detection of anomaly regions in sensor networks with obstacles
    (Datenbanksysteme in Business, Technologie und Web (BTW) – 13. Fachtagung des GI-Fachbereichs "Datenbanken und Informationssysteme" (DBIS), 2009) Franke, Conny; Karnstedt, Marcel; Klan, Daniel; Gertz, Michael; Sattler, Kai -Uwe; Kattanek, Wolfram
    In the past couple of years, sensor networks have evolved to a powerful infrastructure component for monitoring and tracking events and phenomena in many application domains. An important task in processing streams of sensor data is the detection of anoma
  • Konferenzbeitrag
    No Mayfly: Detection and Analysis of Long-term Twitter Trends
    (BTW 2023, 2023) Ziegler, John; Gertz, Michael
    The focus of social media is characterized by stories about short-lived breaking news. Often, such mayflies make it hard to keep track of more profound topics that are prevalent over a longer period of time. To tackle this issue we present a method to detect such long-term trends based on temporal networks and community evolution. Connecting those methods with that of trend analysis allows to study the temporal development of trends"
  • Konferenzbeitrag
    On the State of German (Abstractive) Text Summarization
    (BTW 2023, 2023) Aumiller, Dennis; Fan, Jing; Gertz, Michael
    With recent advancements in the area of Natural Language processing, the focus is slowly shifting from a purely English-centric view towards more language-specific solutions, including German.Especially practical for businesses to analyze their growing amount of textual data are text summarization systems, which transform long input documents into compressed and more digestible summary texts.In this work, we assess the particular landscape of German abstractive text summarization and investigate the reasons why practically useful solutions for abstractive text summarization are still absent in industry. Our focus is two-fold, analyzing a) training resources, and b) publicly available summarization systems.We are able to show that popular existing datasets exhibit crucial flaws in their assumptions about the original sources, which frequently leads to detrimental effects on system generalization and evaluation biases. We confirm that for the most popular training dataset, MLSUM, over 50% of the training set is unsuitable for abstractive summarization purposes. Furthermore, available systems frequently fail to compare to simple baselines, and ignore more effective and efficient extractive summarization approaches. We attribute poor evaluation quality to a variety of different factors, which are investigated in more detail in this work:A lack of qualitative (and diverse) gold data considered for training, understudied (and untreated) positional biases in some of the existing datasets, and the lack of easily accessible and streamlined pre-processing strategies or analysis tools. We therefore provide a comprehensive assessment of available models on the cleaned versions of datasets, and find that this can lead to a reduction of more than 20 ROUGE-1 points during evaluation. As a cautious reminder for future work, we finally highlight the problems of solely relying on n-gram based scoring methods by presenting particularly problematic failure cases. Code for dataset filtering and reproducing results can be found online: https://github.com/anonymized-user/anonymized-repository
  • Konferenzbeitrag
    Online hot spot prediction in road networks
    (Datenbanksysteme für Business, Technologie und Web (BTW), 2011) Häsner, Maik; Junghans, Conny; Sengstock, Christian; Gertz, Michael
    Advancements in GPS-technology have spurred major research and development activities for managing and analyzing large amounts of position data of mobile objects. Data mining tasks such as the discovery of movement patterns, classification and outlier detection in the context of object trajectories, and the prediction of future movement patterns have become basic tools in extracting useful information from such position data. Especially the prediction of future movement patterns of vehicles, based on historical or recent position data, plays an important role in traffic management and planning. In this paper, we present a new approach for the online prediction of so-called hot spots, that is, components of a road network such as intersections that are likely to experience heavy traffic in the near future. For this, we employ an efficient path prediction model for vehicle movements that only utilizes a few recent position data. Using an aggregation model for hot spots, we show how regional information can be derived and connected substructures in a road network can be determined. Utilizing the behavior of such hot spot regions over time in terms of movement or growth, we introduce different types of hot spots and show how they can be determined online. We demonstrate the effectiveness of our approach using a real large-scale road network and different traffic simulation scenarios.