Logo des Repositoriums

Auflistung nach Schlagwort "Information Extraction"

1 - 4 von 4
  • Workshopbeitrag
    Human in the loop information extraction increases efficiency and trust
    (Mensch und Computer 2022 - Workshopband, 2022) Schleith, Johannes; Hoffmann, Hella; Norkute, Milda; Cechmanek, Brian
    Automation is often focused on data-centred measures of success, such as accuracy of the automation or efficiency gain of individual automated steps. This case study shows how a human-assisted information extraction system, that keeps the human in the loop throughout the creation of information extraction rules and their application, can outperform less transparent information extraction systems in terms of overall end-to-end time-on-task and perceived trust. We argue that the time gained through automation can be wiped out by the perceived need of end users to review and comprehend results, where the systems seem obscure to them.
  • Konferenzbeitrag
    MLProvLab: Provenance Management for Data Science Notebooks
    (BTW 2023, 2023) Kerzel, Dominik; König-Ries, Birgitta; Sheeba, Samuel
    Computational notebooks are a form of computational narrative fostering reproducibility.They provide an interactive computing environment where users can run and modify code, and repeat the exploration, providing an iterative communication between data scientists and code. While the ability to execute notebooks non-linearly benefits data scientists for exploration, the drawback is, that it is possible to lose control over the datasets, variables, and methods defined in the notebook and their dependencies.Thus, in this process of user interaction and exploration, there can be a loss of execution history information. To prevent this, a possibility is needed to maintain provenance information. Provenance plays a significant role in data science, especially facilitating the reproducibility of results.To this end, we developed a provenance management tool to help data scientists track, capture, compare, and visualize provenance information in notebook code environments.We conducted an evaluation with data scientists, where participants were asked to find specific provenance information from the execution history of a machine learning Jupyter notebook.The results from the performance and user evaluation show promising aspects of provenance management features of the tool.The resulting system, MLProvLab, is available as an open-source extension for JupyterLab.
  • Zeitschriftenartikel
    Searching-Tool für Compliance
    (HMD Praxis der Wirtschaftsinformatik: Vol. 56, No. 5, 2019) Hengartner, Urs
    Text ist immer noch die vorherrschende Kommunikationsform der heutigen Geschäftswelt. Techniken des Textverstehens erschliessen vielfältiges Wissen zur Verbesserung der Kommunikation zwischen Menschen und Maschinen. In der letzten Zeit haben das automatische Textverstehen und die Extraktion von Semantik bedeutende Fortschritte gemacht. Der Vorteil der Nutzung eines Textanalysesystems für die Überprüfung der Regelkonformität in der Finanzbranche, ist angesichts des Wachstums der Online-Informationen wichtiger denn je. Es ist eine Herausforderung, aktuelle Informationen über Kunden, Unternehmen und Lieferanten zu verfolgen und zu interpretieren. Bei fehlerhaftem Verhalten sind die Auswirkungen auf ein Unternehmen unter Umständen drastisch. Zum Beispiel sind Kundeneröffnungen wegen verordneten Abklärungen für Finanzinstitute oft komplex und kostenintensiv. Um zum Beispiel Missbräuche (Geldwäsche) aufzudecken müssen grosse Mengen an textueller Daten interpretiert werden. Vorgestellt wird ein Anwendungsfall aus der Praxis mit dem Analysewerkzeug Person-Check und den dabei angewandten Textanalysen. Person-Check ermöglicht deutlich effizientere Abklärungen in Compliance-Prüfprozessen unter Berücksichtigung internationaler, lokaler und firmeninternen Richtlinien. Text is still the predominant form of communication in today’s business world. Techniques of text comprehension open up a wide range of knowledge for improving communication between people and machines. Recently, automatic text comprehension and the extraction of semantics have made significant progress. The advantage of using a text analysis system to verify compliance in the financial industry is more important than ever given the growths of online information. It is a challenge to track and interpret current information about customers, companies and suppliers. If an organization behaves incorrectly, the impact can be very drastic. For example, customer openings today are often complex and costly for financial institutions due to mandated clarifications. In order to detect abuses (money laundering), large amounts of textual data must be interpreted. A case study from practice with the textual analysis tool Person-Check and the applied text analytics, will be presented. Person-Check enables significantly more efficient clarifications in compliance audit processes, taking into account international, local and internal company guidelines.
  • Konferenzbeitrag
    Social Relation Extraction from Chatbot Conversations: A Shortest Dependency Path Approach
    (SKILL 2019 - Studierendenkonferenz Informatik, 2019) Glas, Markus
    Digital dialog systems, also known as chatbots, often lack in the sense of a human-like and individualized interaction. The ability to learn someoneŠs social relations during conversations can lead to more personal responses and therefore to a more human-like and diverse conversation. In this work we present S-REX, a comparison method for extracting social relations from chatbot conversations. The implemented approach uses information from the shortest dependency path in combination with state-of-the-art natural language processing models for entity recognition and semantic word vectors. The method is evaluated on two conversational datasets and achieves results close to more complex neural network methods without the need of extensive training.