This thesis develops a system for automatically analyzing and improving dynamic programs, such as those that have driven progress in natural language processing and computer science, more generally, for decades. Finding a correct program with the optimal asymptotic runtime can be unintuitive, time-consuming, and error-prone. This thesis aims to automate this laborious process. To this end, we develop an approach based on 1. a high-level, domain-specific language called Dyna for concisely specifying dynamic programs 2. a general-purpose solver to efficiently execute these programs 3. a static analysis system that provides type analysis and worst-case time/space complexity analyses 4. a rich collection of meaning-preserving transformations to programs, which systematizes the repeated insights of numerous authors when speeding up algorithms in the literature 5. a search algorithm for identifying a good sequence of transformations that reduce the runtime complexity, given an initial, correct program We show that, in practice, automated search -- like the mental search performed by human programmers -- can find substantial improvements to the initial program. Empirically, we show that many speed-ups described in the NLP literature could have been discovered automatically by our system. We provide a freely available prototype system at https://github.com/timvieira/dyna-pi.
Tristan Kneisel, Marko Schmellenkamp, Fabian Vehlken
et al.
This paper explores how natural-language descriptions of formal languages can be compared to their formal representations and how semantic differences can be explained. This is motivated from educational scenarios where learners describe a formal language (presented, e.g., by a finite state automaton, regular expression, pushdown automaton, context-free grammar or in set notation) in natural language, and an educational support system has to (1) judge whether the natural-language description accurately describes the formal language, and to (2) provide explanations why descriptions are not accurate. To address this question, we introduce a representation language for formal languages, Nile, which is designed so that Nile expressions can mirror the syntactic structure of natural-language descriptions of formal languages. Nile is sufficiently expressive to cover a broad variety of formal languages, including all regular languages and fragments of context-free languages typically used in educational contexts. Generating Nile expressions that are syntactically close to natural-language descriptions then allows to provide explanations for inaccuracies in the descriptions algorithmically. In experiments on an educational data set, we show that LLMs can translate natural-language descriptions into equivalent, syntactically close Nile expressions with high accuracy - allowing to algorithmically provide explanations for incorrect natural-language descriptions. Our experiments also show that while natural-language descriptions can also be translated into regular expressions (but not context-free grammars), the expressions are often not syntactically close and thus not suitable for providing explanations.
Anastasia Mavridou, Marie Farrell, Gricel Vázquez
et al.
Integrating autonomous and adaptive behavior into software-intensive systems presents significant challenges for software development, as uncertainties in the environment or decision-making processes must be explicitly captured. These challenges are amplified in safety- and mission-critical systems, which must undergo rigorous scrutiny during design and development. Key among these challenges is the difficulty of specifying requirements that use probabilistic constructs to capture the uncertainty affecting these systems. To enable formal analysis, such requirements must be expressed in precise mathematical notations such as probabilistic logics. However, expecting developers to write requirements directly in complex formalisms is unrealistic and highly error-prone. We extend the structured natural language used by NASA's Formal Requirement Elicitation Tool (FRET) with support for the specification of unambiguous and correct probabilistic requirements, and develop an automated approach for translating these requirements into logical formulas. We propose and develop a formal, compositional, and automated approach for translating structured natural-language requirements into formulas in probabilistic temporal logic. To increase trust in our formalizations, we provide assurance that the generated formulas are well-formed and conform to the intended semantics through an automated validation framework and a formal proof. The extended FRET tool enables developers to specify probabilistic requirements in structured natural language, and to automatically translate them into probabilistic temporal logic, making the formal analysis of autonomous and adaptive systems more practical and less error-prone.
Similar to LLMs, the development of vision language models is mainly driven by English datasets and models trained in English and Chinese language, whereas support for other languages, even those considered high-resource languages such as German, remains significantly weaker. In this work we present an analysis of open-weight VLMs on factual knowledge in the German and English language. We disentangle the image-related aspects from the textual ones by analyzing accu-racy with jury-as-a-judge in both prompt languages and images from German and international contexts. We found that for celebrities and sights, VLMs struggle because they are lacking visual cognition of German image contents. For animals and plants, the tested models can often correctly identify the image contents ac-cording to the scientific name or English common name but fail in German lan-guage. Cars and supermarket products were identified equally well in English and German images across both prompt languages.
This paper investigates the optimal use of the multilingual encoder model mDeBERTa for tasks in three Germanic languages -- German, Swedish, and Icelandic -- representing varying levels of presence and likely data quality in mDeBERTas pre-training data. We compare full fine-tuning with the parameter-efficient fine-tuning (PEFT) methods LoRA and Pfeiffer bottleneck adapters, finding that PEFT is more effective for the higher-resource language, German. However, results for Swedish and Icelandic are less consistent. We also observe differences between tasks: While PEFT tends to work better for question answering, full fine-tuning is preferable for named entity recognition. Inspired by previous research on modular approaches that combine task and language adapters, we evaluate the impact of adding PEFT modules trained on unstructured text, finding that this approach is not beneficial.
Der vorliegende Beitrag ist dem Drama Heldenplatz von Thomas Bernhard gewidmet, seinem wohl bekanntesten Theaterstück, dessen Premiere im Wiener Burgtheater 1988 einen großen Skandal auslöste. Er entzündete sich an den auf der Bühne fallenden abfälligen Worten über Österreich und die Österreicher, die man entweder als eine unzulässige Zumutung oder als eine längst überfällige Abrechnung mit der verdrängten nationalsozialistischen Vergangenheit des Landes interpretierte. Die Analyse des Textes führt zum Ergebnis, dass es Bernhard gar nicht um die vordergründige plakative Österreichkritik ging. Sein eigentliches Anliegen lag vielmehr in der hier benutzten Sprache, deren menschenverachtende Rhetorik auf die ideologischen Raster der NS-Propaganda rekurrierte. Die öffentliche Erregung nach der Premiere zeigte, dass man eine solche Sprache der österreichischen Öffentlichkeit immer noch zumuten konnte. Der Schriftsteller stellte damit eine durchaus beängstigende Diagnose, die die österreichischen Medien und die Diskurskultur des ganzen Landes betraf.
Germanic languages. Scandinavian languages, German literature
Bias studies on multilingual models confirm the presence of gender-related stereotypes in masked models processing languages with high NLP resources. We expand on this line of research by introducing Filipino CrowS-Pairs and Filipino WinoQueer: benchmarks that assess both sexist and anti-queer biases in pretrained language models (PLMs) handling texts in Filipino, a low-resource language from the Philippines. The benchmarks consist of 7,074 new challenge pairs resulting from our cultural adaptation of English bias evaluation datasets, a process that we document in detail to guide similar forthcoming efforts. We apply the Filipino benchmarks on masked and causal multilingual models, including those pretrained on Southeast Asian data, and find that they contain considerable amounts of bias. We also find that for multilingual models, the extent of bias learned for a particular language is influenced by how much pretraining data in that language a model was exposed to. Our benchmarks and insights can serve as a foundation for future work analyzing and mitigating bias in multilingual models.
Tanja Angela Kunz, Matthias Buschmeier, Jens Ciecior
et al.
Die gesamtgesellschaftlichen Umstellungsprozesse im Kontext von Digitalisierung und Digitisierung haben längst und umfassend auch die traditionelle Konzentration auf den papiernen Text als Gegenstand geisteswissenschaftlichen Studierens und Forschens sowie die Anleitung von Lern- und Rezeptionsprozessen in der Präsenzlehre erfasst. Dieser Transformationsprozess ist durch die Covid-19-Pandemie weniger erzeugt denn intensiviert worden. Augenfällig ist in jedem Fall der Bedarf an fachwissenschaftlich und technisch professionell aufbereiteten digitalen Lehr‐Lern‐Materialien, um in diesem Wandel die wissenschaftlichen Standards akademischer Lehre zu behaupten und weiter zu verbessern. Der Beitrag zeigt am Beispiel des Modellprojekts KoLidi (Kollaborative Literaturgeschichte digital und interaktiv) der Universitäten Bielefeld, Paderborn und der Bergischen Universität Wuppertal die Möglichkeiten digitalen Lehren und Lernens innerhalb eines traditionell buchzentrierten und lektüreintensiven Fachs auf. Das Projekt stellte einen breit gelagerten Versuch
zur hochschuldidaktisch und fachwissenschaftlich qualifizierten Aufbereitung von Seminaren im digitalen Raum auf Basis der technischen Möglichkeiten von Lernmanagementsystemen (LMS) dar. Im Zentrum des Beitrags stehen die digitale kollaborative und interaktive Gestaltung von Rezeptions‐ und Textproduktionsprozessen sowie die in digitale Formate ‚übersetzte‘ Lese‐ und Schreibbegleitung.
Abstract (english):
Processes of digitization and digitalization fundamentally transform our societies. More precisely, they have long since comprehensively impacted the traditional focus on paper-printed texts as a commodity of studies and research in the humanities. Furthermore, they have largely affected learning and reception processes in faceto-face classroom situations. Less has this transformation been generated by the
Covid-19-pandemic than it has become visible through it. Needs for professionally prepared digital teaching and learning materials have become obvious, in order to maintain and further improve the scientific standards of academic teaching in digital times. For disciplinary subjects, like literary history, that traditionally focus on books and large amounts of more or less analogue readings this has become an urgent challenge. Using examples from the pilot scheme KoLidi (‘Kollaborative Literaturgeschichte digital und interaktiv’, which roughly translates into ‘Collaborative and Interactive Approaches for Digital Literary History), this article presents well working ways of digital mediation of German-language literary history. The project was carried out conjointly by scholars from the universities of Bielefeld, Paderborn and Wuppertal. It is an effort to set up a didactically and technically qualified digital working environment and study materials at university level within a Moodle-based learning management system. The paper focuses, on the one hand, on digital, collaborative and interactive processes of reception and textual production as well as their designs. On the other hand, it discusses how reading and writing scenarios can be converted into digital formats and how students are best supported in reading and writing processes.
The present study examines the use of the hybrid learning model in lessons for teaching German as a foreign language carried out at a Turkish university in the 2021-2022 preparatory year. The quantitative investigation aims to evaluate the model that was used from students’ perspectives to gain insight regarding discussions on future hybrid learning models. Data were collected using a semi-structured, non-standardized online questionnaire with Likert-type statements and open ended questions. As a result, the study determined the opinions of 61 students regarding motivation and learning efficiency for the online and face-to-face aspects of the lesson, the features of hybrid learning, and the likelihood of its continuation. The data were evaluated numerically using Google Forms and Microsoft Excel. The results show that the students rated their motivation and learning efficiency significantly higher regarding the face-to-face aspects of the lesson compared to the online aspects, which can certainly be explained by the adverse effects of the COVID-19 pandemic. Evaluating the general Likert-type statements regarding hybrid learning revealed the students’ attitudes toward several aspects of hybrid learning. For example, the students stated that the digital activities used for the online parts of the lesson (e.g., interactive exercises) are able to support foreign language learning and that the face-to-face parts of the lesson positively emphasized collaborative learning in a community. When asked about the likelihood of continuing hybrid learning, the students turned out to be half in favor and half opposed. Due to the results from the present study, pedagogical added value should be placed more at the center of all didactic considerations when designing future hybrid learning classes.
German literature, Germanic languages. Scandinavian languages
This paper examines approaches to generate lexical resources for endangered languages. Our algorithms construct bilingual dictionaries and multilingual thesauruses using public Wordnets and a machine translator (MT). Since our work relies on only one bilingual dictionary between an endangered language and an "intermediate helper" language, it is applicable to languages that lack many existing resources.
Norwegian Twitter data poses an interesting challenge for Natural Language Processing (NLP) tasks. These texts are difficult for models trained on standardized text in one of the two Norwegian written forms (Bokmål and Nynorsk), as they contain both the typical variation of social media text, as well as a large amount of dialectal variety. In this paper we present a novel Norwegian Twitter dataset annotated with POS-tags. We show that models trained on Universal Dependency (UD) data perform worse when evaluated against this dataset, and that models trained on Bokmål generally perform better than those trained on Nynorsk. We also see that performance on dialectal tweets is comparable to the written standards for some models. Finally we perform a detailed analysis of the errors that models commonly make on this data.
Andrea Padovan, Ermenegildo Bidese, Alessandra Tomaselli
In our paper, we deal with the Germanic–Romance language contact, focusing on Cimbrian, a Germanic minority language spoken in Northern Italy. Specifically, we focus on the violation of the well-known that-trace filter, as it appears to be an interesting case of the superficial convergence that we ascribe to the status of T, which is either too rich (model language) or too weak (replica language) to represent a viable landing site for subject extraction.
Angesichts der Tatsache, dass Schüler*innen digitale Medien mit großer Selbstver-ständlichkeit nutzen und deren Bedeutung aufgrund ihrer Verwobenheit mit lebens-weltlicher Erfahrung möglicherweise gar nicht mehr in Frage stellen, untersucht der Beitrag, wie das fachliche Lernen im inklusiven Literaturunterricht angesichts sich verändernder Kommunikations- und Rezeptionsformen im Zeitalter der Digitalität gestaltet werden kann. Am Beispiel von Volker Kutschers Kriminalroman Der nasse Fisch und seinen medialen Dispositiven wird im Folgenden für ein reflexives Ver-ständnis kultureller Teilhabe im inklusiven Deutschunterricht plädiert. Davon ausgehend werden Überlegungen für die Umsetzung im inklusiven Deutschunterricht vorgestellt, die eine Reflexion digitaler Transformationsprozesse des Erzählens im Sinne einer Critical Narrative Literacy fokussiert.
Abstract (english): Inclusive Literature Teaching in the Digital Age – Reflections on Kutscher‘s Crime Novel Der nasse Fisch and its Media Dispositives
With regard to the fact that students use digital media in a most natural fashion and may even no longer question their significance due to their interconnectedness with worldly experience, the article examines how subject-specific learning can be shaped in view of changing forms of communication and reception in the digital age. Using the example of Volker Kutscher‘s crime novel Der nasse Fisch and its media dispositives, the following article argues for a reflexive comprehension of cultural participation in inclusive learning environments. Based on this, some ideas for settings in inclusive classrooms are presented which focus on reflecting the digital transformation processes of narratives in terms of Critical Narrative Literacy.
Somnath Banerjee, Maulindu Sarkar, Nancy Agrawal
et al.
Hate speech is considered to be one of the major issues currently plaguing online social media. Repeated and repetitive exposure to hate speech has been shown to create physiological effects on the target users. Thus, hate speech, in all its forms, should be addressed on these platforms in order to maintain good health. In this paper, we explored several Transformer based machine learning models for the detection of hate speech and offensive content in English and Indo-Aryan languages at FIRE 2021. We explore several models such as mBERT, XLMR-large, XLMR-base by team name "Super Mario". Our models came 2nd position in Code-Mixed Data set (Macro F1: 0.7107), 2nd position in Hindi two-class classification(Macro F1: 0.7797), 4th in English four-class category (Macro F1: 0.8006) and 12th in English two-class category (Macro F1: 0.6447).