A5072459546- Natural Language Processing Techniques
- Topic Modeling
- Text Readability and Simplification
- Advanced Text Analysis Techniques
- Semantic Web and Ontologies
- Wikis in Education and Collaboration
- Hate Speech and Cyberbullying Detection
- Sentiment Analysis and Opinion Mining
- Speech Recognition and Synthesis
- Handwritten Text Recognition Techniques
- Speech and dialogue systems
- Web Data Mining and Analysis
- Biomedical Text Mining and Ontologies
- Text and Document Classification Technologies
- Software Engineering Research
- Image Processing and 3D Reconstruction
- Intelligent Tutoring Systems and Adaptive Learning
- Educational Technology and Assessment
- Digital Communication and Language
- Authorship Attribution and Profiling
- Second Language Acquisition and Learning
- Adversarial Robustness in Machine Learning
- Music and Audio Processing
- Educational Assessment and Pedagogy
- Hand Gesture Recognition Systems
University of Hagen
2022-2025
Innovation Cluster (Canada)
2022
University of Duisburg-Essen
2014-2021
Dartmouth College
2021
University of Stuttgart
2021
Stockholm University
2021
Uppsala University
2021
East Stroudsburg University
2021
DIPF | Leibniz Institute for Research and Information in Education
2012-2016
Educational Testing Service
2015
Neural approaches to automated essay scoring have recently shown state-of-the-art performance. The task typically involves a broad notion of writing quality that encompasses content, grammar, organization, and conventions. This differs from the short answer content task, which focuses on accuracy. inputs neural models – ngrams embeddings are arguably well-suited evaluate in tasks. We investigate how several basic similar those used for perform scoring. show architectures can outperform...
Abstract In this article, we present a comprehensive study aimed at computing semantic relatedness of word pairs. We analyze the performance large number measures proposed in literature with respect to different experimental conditions, such as (i) datasets employed, (ii) language (English or German), (iii) underlying knowledge source, and (iv) evaluation task (computing scores relatedness, ranking pairs, solving choice problems). To our knowledge, is first systematically on properties,...
Automated scoring of student essays is increasingly used to reduce manual grading effort.State-of-the-art approaches use supervised machine learning which makes it complicated transfer a system trained on one task another.We investigate currently features are task-independent and evaluate their transferability English German datasets.We find that, by using our feature set, models better between tasks.We also that the works even tasks same type.
Language proficiency tests are used to evaluate and compare the progress of language learners. We present an approach for automatic difficulty prediction C-tests that performs on par with human experts. On basis detailed analysis newly collected data, we develop a model C-test introducing four dimensions: solution difficulty, candidate ambiguity, inter-gap dependency, paragraph difficulty. show cues from all dimensions contribute
We present DKPro TC, a framework for supervised learning experiments on textual data.The main goal of TC is to enable researchers focus the actual research task behind problem and let handle rest.It enables rapid prototyping by relying an easy-to-use workflow engine standardized document preprocessing based Apache Unstructured Information Management Architecture (Ferrucci Lally, 2004).It ships with standard feature extraction modules, while at same time allowing user add customized...
Abstract This paper introduces MultiGEC, a dataset for multilingual Grammatical Error Correction (GEC) in twelve European languages: Czech, English, Estonian, German, Greek, Icelandic, Italian, Latvian, Russian, Slovene, Swedish and Ukrainian. MultiGEC distinguishes itself from previous GEC datasets that it covers several underrepresented languages, which we argue should be included resources used to train models Natural Language Processing tasks which, as itself, have implications Learner...
Semantic relatedness is a special form of linguistic distance between words. Evaluating semantic measures usually performed by comparison with human judgments. Previous test datasets had been created analytically and were limited in size. We propose corpus-based system for automatically creating datasets. Experiments subjects show that the resulting cover all degrees relatedness. As result approach, types lexical-semantic relations contain domain-specific words naturally occurring texts.
Automatic content scoring is an important application in the area of automatic educational assessment. Short texts written by learners are scored based on their while spelling and grammar mistakes usually ignored. The difficulty automatically such varies with variance within learner answers. In this paper, we first discuss factors that influence answers, so practitioners can better estimate if might be applicable to usage scenario. We then compare two main paradigms scoring: (i)...
Automated short answer scoring is increasingly used to give students timely feedback about their learning progress.Building models comes with high costs, as stateof-the-art methods using supervised require large amounts of hand-annotated data.We analyze the potential recently proposed for semi-supervised based on clustering.We find that all examined (centroids, clusters, selected pure clusters) are mainly effective very answers and do not generalize well severalsentence responses.
Automatically generating challenging distractors for multiple-choice gap-fill items is still an unsolved problem.We propose to employ context-sensitive lexical inference rules in order generate that are semantically similar the gap target word some sense, but not particular sense induced by context.We hypothesize such should be particularly hard distinguish from correct answer.We focus on verbs as they especially difficult master language learners and find our approach quite effective.In...
Lexical recognition tests are widely used to assess vocabulary knowledge. We investigate the role that diacritics play in designing an Arabic lexical test. compare a non-diacritized and diacritized test user study find they largely comparable their ability proficiency. However, we argue better suited control difficulty by allowing nonwords more targeted selection of word forms.