OPUS - Publikationenserver der Technischen Hochschule Nürnberg Georg Simon Ohm
Not a member yet
2332 research outputs found
Sort by
A Multimodal Approach to Device-Directed Speech Detection with Large Language Models
No full textInteractions with virtual assistants typically start with a predefined trigger phrase followed by the user command. To make interactions with the assistant more intuitive, we explore whether it is feasible to drop the requirement that users must begin each command with a trigger phrase. We explore this task in three ways: First, we train classifiers using only acoustic information obtained from the audio waveform. Second, we take the decoder outputs of an automatic speech recognition (ASR) system, such as 1-best hypotheses, as input features to a large language model (LLM). Finally, we explore a multimodal system that combines acoustic and lexical features, as well as ASR decoder signals in an LLM. Using multimodal information yields relative equal-error-rate improvements over text-only and audio-only models of up to 39% and 61%. Increasing the size of the LLM and training with low-rank adaption leads to further relative EER reductions of up to 18% on our dataset
Towards Self-Attention Understanding for Automatic Articulatory Processes Analysis in Cleft Lip and Palate Speech
No full textCleft lip and palate (CLP) speech presents unique challenges for automatic phoneme analysis due to its distinct acoustic characteristics and articulatory anomalies. We perform phoneme analysis in CLP speech using a pre-trained wav2vec 2.0 model
with a multi-head self-attention classification module to capture long-range dependencies within the speech signal, thereby enabling better contextual understanding of phoneme sequences. We demonstrate the effectiveness of our approach in the classification of various articulatory processes in CLP speech. Furthermore, we investigate the interpretability of self-attention to gain insights into the model’s understanding of CLP speech characteristics. Our findings highlight the potential of the selfattention mechanisms for improving automatic phoneme analysis in CLP speech, paving the way for enhanced diagnostics, adding interpretability for therapists and affected patients
Storytelling im Change-Prozess
No full textDer Journalismus und die Werbebrache haben es vorgemacht. Seit kurzem setzt auch Change Management verstärkt auf Storytelling, um die menschliche Seite bei Veränderungen zu begleiten. Die Protagnisten in den authentischen Geschichten sind meist Kolleginnen und Kollegen, mit denen sich die vom Wandel betroffenen Mitarbeitenden am besten identifizieren können. Die Ausspielkanäle bzw. Kommunikationsformen reichen vom Intranet und Mitarbeiterzeitungen über Podcasts und Videos bis hin zu Workshops und Personalversammlungen. Damit eine Geschichte im Rahmen der Change Communication wirkt und glaubwürdig ist, darf sie wie im Journalismus nicht erfunden sein. Sie ähnelt daher häufig den journalistischen Darstellungsformen Reportage und Porträt. Um diese zu verfassen, sind journalistische Vorerfahrungen für Change Manager von unschätzbarem Vorteil
Look What’s There! Utilizing the Internet’s Existing Data for Censorship Circumvention with OPPRESSION
No full textAn ongoing challenge in censorship circumvention is optimizing the stealthiness of communications, enabled by covert channels. Recently, a new variant called history covert channels has been proposed. Instead of modifying or mimicking legitimate data, such channels solely point to observed data matching secret information. This approach reduces the amount of secret data a sender explicitly must transfer and thus limits detectability. However, the only published history channel is only suitable for special scenarios due to severe limitations in terms of bandwidth. We propose a significant
performance enhancement of history covert channels that allows their use in real-world scenarios through utilizing the content of online social media and online archives. Our approach, which we call OPPRESSION (Open-knowledge Compression), takes advantage
of the massive amounts of textual data on the Internet that can be referenced by short pointer messages. Broadly, OPPRESSION can be considered a novel encoding strategy for censorship circumvention. We further present and evaluate our open source proof-of-concept implementation of OPPRESSION that can transfer secret data by pointing to popular online media, such as Twitter (now “X”), news websites, Wikipedia entries, and online books. The pointer itself is transmitted through existing censorship circumvention systems. Our approach minimizes the amount of traffic to be concealed in comparison to existing works, even in comparison to compression
Unraveling the gender wage gap: Exploring early career patterns among university graduates
No full textAbstractA large body of literature has shown that the gender wage gap is small in the first years after graduation and increases gradually with age, largely because of family decisions, often a penalty caused by childbirth. However, the gender wage gap immediately after graduation has received less attention. Using a unique dataset that links 5000 university graduates with master's degrees or equivalent from a large German university to detailed employment records from the German social security register, we specifically analyze the gender wage gap at the first job and its dynamics during the initial years of their careers after graduation. We find that a significant gender wage gap already exists in the first job after graduation, even before most young individuals make family decisions. However, this gender wage gap decreases in the first year after entering the labor market and then increases slowly over time. We attribute this initial decrease in the gender wage gap to female university graduates experiencing greater returns from firm and occupational changes than their male counterparts. This suggests that women may use these changes to address skill mismatches, which are more common among women than men in their first job
Change in der Medien- und Kommunikationsbranche
No full textMit dem vorliegenden Band erhalten Medienschaffende, Digitalisierungsmanager in lokalen und regionalen Medienhäusern und nicht zuletzt Studierende der Medien- und Kommunikationsbranche einen Praxisleitfaden, um die Digitalisierung zu gestalten und den "Wind of Change" in eine erfolgreiche, vielfalterhaltende und damit demokratiesichernde Richtung zu lenken. Es wird neben einer fundierten Einführung in Change Management auch das Thema Leadership behandelt sowie die neue Rolle der Chefredakteurinnen und Chefredakteure. Außerdem werden neue Berufsbilder in der Medienbranche vorgestellt
Einsatz, Nutzen und Grenzen von ChatGPT und anderen Large Language Modellen an den bayerischen HAWs
Get PDFDie Studie analysiert Einsatz, Nutzen und Grenzen von ChatGPT und Large Language Modellen (LLMs) in der Lehre an bayerischen Hochschulen für angewandte Wissenschaften (HAWs). Sie basiert auf einer schriftlichen Onlinebefragung und leitfadengestützten Interviews unter Lehrenden, Studierenden und Funktions-trägerinnen und -träger. Insgesamt nahmen 1570 Personen an der Onlineumfrage teil und es wurden 27 Interviews geführt. Zwar variieren die Ergebnisse je nach Befragtengruppe, aber tendenziell lassen sich folgende Einschätzungen auf Basis der empirischen Daten treffen:
→ Die meisten Teilnehmenden haben ein geringes Verständnis der Nutzbarkeit von LLMs und deren technologischer Funktionsweise. Die Nutzbarkeit von Open Source LLMs oder domänenspezifisch trainierten LLMs für das eigene Fachgebiet kennt der Großteil der Teilnehmenden (überhaupt) nicht oder nur teilweise.
→ Die Einsatzhäufigkeit von ChatGPT bzw. LLMs in Lehre, Studium und Arbeitsalltag ist noch gering, vor allem wegen fehlender sinnvoller Einsatzmöglichkeiten, mangelnder Qualität der Ergebnisse und rechtlicher Bedenken. Die ChatGPT-Plusversion wird von den meisten Teilnehmenden nie genutzt.
→ Die wahrgenommene Nützlichkeit der Einsatzszenarien von ChatGPT variiert je nach ChatGPT-Version und Zielgruppe, wobei das Zusammenfassen und die Verbesserung von Texten als nützlich eingeschätzt werden.
→ Ein Verbot von ChatGPT wird überwiegend abgelehnt. Stattdessen besteht seitens der Lehrenden und Studierenden der Wunsch nach vermehrter Aufklärung über die Chancen und Grenzen von ChatGPT und Co. Zudem erhoffen sie sich finanzielle, technische und didaktische Unterstützung durch die Hochschulen.
→ Die Teilnehmenden sehen ChatGPT als eine Chance zur Verbesserung der Lehre, aber auch als eine Herausforderung für die Prüfungskultur, die rechtliche Sicherheit und die Kompetenzentwicklung.
• In Bezug auf zukünftig notwendige Kompetenzen wird angemerkt, dass es zu einem Kompetenz-verlust bei den Studierenden als Konsequenz der ChatGPT-Nutzung kommen kann.
• Die Frage nach der rechtlichen Sicherheit bezieht sich nicht nur auf mögliche Datenschutz- oder Copyrightverletzungen, sondern auch auf das Prüfungswesen.
• Die Art und Weise von Prüfungen wird sich ändern (müssen), wobei mehrere Optionen thematisiert werden: von der Kennzeichnungspflicht bei der Nutzung von ChatGPT und Co. über den Wegfall reiner Wissensabfragen in Prüfungen bis hin zu den Ideen, keine theoretischen Bachelorarbeiten mehr zu vergeben und vermehrt mündliche Prüfungen einzusetzen.
Die Studie schließt mit einer Diskussion über die Rolle und Tragweite der generativen KI in der Hoch-schullehre und gibt Hinweise auf Materialien für die Gestaltung von innovativen und verantwortungs-vollen LernszenarienThe study analyses use, benefits, and limitations of ChatGPT and Large Language Models (LLMs) in teaching at Bavarian universities of applied sciences. It is based on a written online survey and guideline-based interviews with teachers, students, and functionaries. A total of 1,570 individuals participated in the on-line survey, and 27 interviews were conducted. Although results vary depending on the respondent group, the following general trends can be estab-lished on empirical data:
→ Most participants only have limited understanding of the usability of ChatGPT or LLMs and the underlying technology. The majority of participants is not (at all) or only partially familiar with open-source LLMs and LLMs specifically trained for their field.
→ The usage frequency of ChatGPT or LLMs in teaching, studying and daily professional life is still low, mainly due to the lack of meaningful application possibilities, the poor quality of results, and legal concerns. The ChatGPT Plus version is hardly used at all.
→ The perceived usefulness of application scenarios varies by version and target group, with text summarization and text improvement being rated as useful.
→ A ban on ChatGPT is predominantly rejected. Instead, teachers and students want more information about the opportunities and limitations of ChatGPT and similar tools. They are also hoping for financial, technical and didactic support from universities.
→ The participants see ChatGPT as an opportunity to improve teaching, but also as a challenge for examination practices, legality, and competency development.
• Concerning competencies required in the future, it is noted that the use of ChatGPT may lead to a loss of competencies among students.
• The question of legality not only encompasses data protection or copyright violations, but also examination practices.
• The way examinations are conducted will (have to) change, with several options being discussed: from mandatory labelling when ChatGPT and similar tools are used to the elimination of knowl-edge tests to the idea of no longer assigning theoretical Bachelor‘s theses and increasing the use of oral exams.
The study concludes with a discussion of the role and scope of generative AI in higher education teaching and provides insights on materials for designing innovative and responsible learning scenarios
Mechanical properties of carbon nanotube (CNT) reinforced polymers using electron‐deficient aromatics as additives
No full textAbstract In initial experiments the effects of aromatic model substances on carbon nanotube (CNT) dispersions in dimethylformamide (DMF) were investigated. Electron‐deficient aromatics interact strongly with CNTs, causing increased agglomeration and sedimentation. Conversely, electron‐donating aromatics stabilize CNT dispersions in DMF. Polymers with electron‐deficient aromatics, such as polyd initrostyrene (PDNS), exhibit a concentration‐dependent effect: low concentrations lead to stabilization of dispersions, while higher concentrations lead to sedimentation. This suggests that such polymers can enhance attraction between the matrix of CNT‐reinforced polymers as well as stabilize the dispersed CNTs. Polycarbonate, modified with polydinitrocarbonate (PDNC) and reinforced with CNTs showed improved mechanical properties. The addition of 6 wt.% CNTs and 6 wt.% PDNC resulted in a notable improvement with a 22% increase in tensile strength, a 29% increase in flexural strength, a 39% increase in Young's modulus and a 47% increase in flexural modulus. This enhancement resulted in an overall mechanical performance comparable to the high‐performance polymer polyetherimide. However, there must be noted, that the addition of PDNC increases the CNT particle size, which can negatively affect mechanical properties. The results highlight the additive's dual role in enhancing adhesive interactions while potentially increasing CNT agglomerate sizes.Highlights
Interactions of CNTs dispersed in DMF and various aromatics were investigated.
Polydinitrocarbonate (PDNC) was synthesized as a new additive for CNT‐composites.
Polycarbonate/CNT‐composites were obtained using extrusion.
Test specimens with CNT contents up to 6 wt.% were obtained.
Mechanical properties of polycarbonate reached the level of polyetherimide
Modeling Programming Competency
No full textThis book covers a qualitative study on the programming competencies of novice learners in higher education. To be precise, the book investigates the expected programming competencies within basic programming education at universities and the extent to which the Computer Science curricula fail to provide transparent, observable learning outcomes and assessable competencies. The study analyzes empirical data on 35 exemplary universities' curricula and interviews with experts in the field. The book covers research desiderata, research design and methodology, an in-depth data analysis, and a presentation and discussion of results in the context of programming education. Addressing programming competency in such great detail is essential due to the increasing relevance of computing in today’s society and the need for competent programmers who will help shape our future.
Although programming is a core tier of computing and many related disciplines, learning how to program can be challenging in higher education, and many students fail in introductory programming. The book aims to understand what programming means, what programming competency encompasses, and what teachers expect of novice learners. In addition, it illustrates the cognitive complexity of programming as an advanced competency, including knowledge, skills, and dispositions in context. So, the purpose is to communicate the breadth and depth of programming competency to educators and learners of programming, including institutions, curriculum designers, and accreditation bodies. Moreover, the book’s goal is to represent how a qualitative research methodology can be applied in the context of computing education research, as the qualitative research paradigm is still an exception in computing education research.
The book provides new insights into programming competency. It outlines the components of programming competencies in terms of knowledge, skills, and dispositions and their cognitive complexity according to the CC2020 computing curricula and the Anderson-Krathwohl taxonomy of the cognitive domain. These insights are essential as programming constitutes one of the most relevant competencies in all computing study programs. In addition, being able to program describes the capability of solving problems, which is also a core competency in today’s increasingly digitalized society. In particular, the book reveals the great relevance of dispositions and other competency components in programming education, which curricula currently fail to recognize and specify. In addition, the book outlines the resulting implications for higher education institutions, educators, and student expectations. Yet another result of interest to graduate students is the multi-method study design that allows for the triangulation of data and results