Linköping Electronic Conference Proceedings
Not a member yet
1113 research outputs found
Sort by
Out-of-the-Box Graded Vocabulary Lists with Generative Language Models: Fact or Fiction?
Get PDFIn this paper, we explore the zero-shot classification potential of generative language models for the task of grading vocabulary and generating graded vocabulary lists. We expand upon prior research by testing five different language model families on five different languages. Our results indicate that generative models can grade vocabulary across different languages with moderate but stable success, but producing vocabulary in a language other than English seems problematic and often leads to the generation of non-words, or words in a language other than the target language
Investigating strategies for lexical complexity prediction in a multilingual setting using generative language models and supervised approaches
Get PDFThis paper explores methods to automatically predict lexical complexity in a multilingual setting using advanced natural language processing models. More precisely, it investigates the use of transfer learning and data augmentation techniques in the context of supervised learning, showing the great interest of multilingual approaches. We also assess the potential of generative large language models for predicting lexical complexity. Through different prompting strategies (zero-shot, one-shot, and chain-of-thought prompts), we analyze model performance in diverse languages. Our findings reveal that while generative models achieve high correlation scores, their predictive quality varies. The comparative study illustrates that while generative large language models have potential, optimized task-specific models still outperform them in accuracy and reliability
Jingle BERT, Jingle BERT, Frozen All the Way: Freezing Layers to Identify CEFR Levels of Second Language Learners Using BERT
No full textIn this paper, we investigate the question of how much domain adaptation is needed for the task of automatic essay assessment by freezing layers in BERT models. We test our methodology on three different graded language corpora (English, French and Swedish) and find that partially fine-tuning base models improves performance over fully fine-tuning base models, although the number of layers to freeze differs by language. We also look at the effect of freezing layers on different grades in the corpora and find that different layers are important for different grade levels. Finally, our results represent a new state-of-the-art in automatic essay classification for the three languages under investigation
A Conversational Intelligent Tutoring System for Improving English Proficiency of Non-Native Speakers via Debriefing of Online Meeting Transcriptions
Get PDFThis paper presents work-in-progress on developing a conversational system designed to enhance non-native English speakers' language skills through post-meeting analysis of the transcriptions of video conferences in which they have participated. Following recent advances in chatbots and agents based on large language models (LLMs), our tutoring system leverages pre-trained LLMs within an ecosystem that integrates different techniques, including in-context learning, external non-parametric memory retrieval, efficient parameter fine-tuning, grammatical error correction models, and error-preserving speech synthesis. A detailed analysis of the different technologies employed in each of these aspects is provided, along with a description of the datasets used. The system is currently in development, with a planned pilot study to evaluate its effectiveness among students of L2-English
Cognitive Assessment and Profiling for increased understanding of Individual and Team Game Intelligence and Performance in Ice hockey
No full textGame intelligence, the ability to be in the right place at the right time and make optimal decisions, is crucial for athletic performance. This whitepaper explores how neurocognitive testing and profiling can deepen our understanding of game intelligence, which includes elements such as situational awareness, decision-making, problem-solving, and flexibility.
The whitepaper targets sports professionals aiming to enhance their understanding of game intelligence through neurocognitive assessments. The assessments mentioned in the paper provide insights into athletes’ cognitive strengths and weaknesses, aiding in talent identification, personalized coaching, strategic team composition, tactical adaptations, and injury prevention. Executive functions are crucial in both open sports (e.g., soccer, basketball) and closed sports (e.g., archery, golf). For example, in ice hockey, players must continuously adapt to dynamic environments, requiring quick decision-making, strategic thinking, and creativity.
Integrating neurocognitive assessments into sports practices has the potential to enhance the understanding of game intelligence, reduce subjectivity and bias, and improve individual and team performance, As well as ensure the wellbeing of athletes through tailored mental health support and coping strategies. Testing and profiling of individuals and teams can practically help enhance understanding of Game Intelligence. The process involves assessment, awareness, individual acceptance, strategic development, integration into coaching, and continuous follow-up to monitor progress and aid adjustments
Predicting Overtakes In Trucks Using Can Data
Get PDFSafe overtakes in trucks are crucial to prevent accidents, reduce congestion, and ensure efficient traffic flow, making early prediction essential for timely and informed driving decisions. Accordingly, we investigate the detection of truck overtakes from CAN data. Three classifiers, Artificial Neural Networks (ANN), Random Forest, and Support Vector Machines (SVM), are employed for the task. Our analysis covers up to 10 seconds before the overtaking event, using an overlapping sliding window of 1 second to extract CAN features. We observe that the prediction scores of the overtake class tend to increase as we approach the overtake trigger, while the no-overtake class remain stable or oscillates depending on the classifier. Thus, the best accuracy is achieved when approaching the trigger, making early overtaking prediction challenging. The classifiers show good accuracy in classifying overtakes (Recall/TPR ≥ 93%), but accuracy is suboptimal in classifying no-overtakes (TNR typically 80-90% and below 60% for one SVM variant). We further combine two classifiers (Random Forest and linear SVM) by averaging their output scores. The fusion is observed to improve no-overtake classification (TNR ≥ 92%) at the expense of reducing overtake accuracy (TPR). However, the latter is kept above 91% near the overtake trigger. Therefore, the fusion balances TPR and TNR, providing more consistent performance than individual classifiers
Profiles for Swedish as a Second Language: Lexis, Grammar, Morphology
Get PDFThis article gives a short introduction to the Swedish Second Language Profile, a tool that visualizes language in Swedish learner corpora from different angles, such as vocabulary, grammar and morphology. The tool is aimed at research on Second Language Acquisition, development of NLP models, teaching of Swedish as a second language, automatic approaches for second language teaching and learning, and at a number of other fields
From Zipf distribution to Universal Dependencies – Interactive Notebooks for Swedish Text Analysis
Get PDFNotebook-based environments are powerful (web-based) interactive development resources for conducting exploratory (textual) data analysis (EDA). These environments allow the embedding of code (code snippets in ‛code cells’) which can be easily executed with the results immediately presented into the user’s window. This paper introduces some basic exploratory tools and techniques using JupyterLab notebooks, applied to Swedish using a subcorpus that address various topics related to the COVID-19 pandemic published during January-December 2021
DASH Swedish National Doctoral School in Digital Humanities: From Local Expertise to National Research Infrastructure
Get PDFThis paper presents the Swedish National Doctoral School in Digital Humanities: Data, Culture, and Society – Critical Perspectives (DASH) that is run in 2023–2027 by Uppsala University, Umeå University, Linnaeus University, and Gothenburg University. Though Swedish universities have established PhD courses, MA programmes and training in digital humanities previously, DASH is the first encompassing educational programme in digital humanities at the doctoral level. The present paper discusses the rationale behind the DASH doctoral school, its role in the landscape of Swedish humanities infrastructures, and provides insights from the first PhD courses and seminars. The focus of DASH is to equip PhD candidates in humanities and social sciences with knowledge and skills necessary to pursue high quality, innovative and critical research in digital humanities. DASH aims to provide knowledge in relation to digital research, its methods, tools, and critical perspectives, and to build and strengthen the networks among early career scholars. DASH facilitates access and use of the resources in the national infrastructures in the humanities, but also emerges as an element in the infrastructure by providing new resources and competences
Linguistic Autobiographies. Towards the Creation of a Multilingual Resource Family
Get PDFThis paper describes a project aimed at creating a new resource family of multilingual and multimodal resources centered around the concept of “Linguistics of self”, that is personal re-flections on the role of languages in shaping one’s identity. Language portrait silhouettes, drawing bilingualism, and linguistic autobiographies are different types of resources that share this common feature. We describe the resources and the criteria for their metadata annotation, focusing in particular on linguistic autobiographies, where the writer explicitly reflects on the relationship between him/herself and language. These genres are fruitfully used in different educational settings, and research has shown that they help to uncover the social, affective, and psychological dimensions of language learning. The potential of a multilingual and mul-timodal collection is discussed starting from data collected in Italy and Norway