With the advent of Generative AI models, the automatic generation of educational questions plays a key role in developing online education. This work compares large-language model-based (LLM) systems and their small-l...
详细信息
ISBN:
(纸本)9783031643149;9783031643156
With the advent of Generative AI models, the automatic generation of educational questions plays a key role in developing online education. This work compares large-language model-based (LLM) systems and their small-language model (sLM) counterparts for educational question generation. Our experiments, quantitatively and qualitatively, demonstrate that sLMs can produce educational questions with comparable quality by further pre-training and fine-tuning.
In the assessment of essay writing, reliably measuring examinee ability can be difficult owing to bias effects arising from rater characteristics. To address this, item response theory (IRT) models that incorporate ra...
详细信息
ISBN:
(纸本)9783031643118;9783031643125
In the assessment of essay writing, reliably measuring examinee ability can be difficult owing to bias effects arising from rater characteristics. To address this, item response theory (IRT) models that incorporate rater characteristic parameters have been proposed. These models estimate the ability of examinees from scores assigned by multiple raters while considering their scoring characteristics, thereby achieving more accurate measurement of ability compared with a simple average of scores. However, issues arise when different groups of examinees are assessed by distinct sets of raters. In such cases, test linking is required to standardize the scale of ability estimates among multiple examinee groups. Traditional test linking methods require administrators to design groups in which either examinees or raters are partially shared-a requirement that is often impractical in real-world assessment settings. To overcome this problem, we introduce a novel linking method that does not rely on common examinees and raters by utilizing a recent automated essay scoring (AES) method. Our method not only facilitates test linking but also enables effective collaboration between human raters and AES, which enhances the accuracy of ability measurement.
This paper presents a cross-lingual methodology for analyzing verbal argument structures to uncover shared syntax-semantic patterns among verbal complements across languages. The primary contribution is a novel semant...
详细信息
ISBN:
(纸本)9783031659898;9783031659904
This paper presents a cross-lingual methodology for analyzing verbal argument structures to uncover shared syntax-semantic patterns among verbal complements across languages. The primary contribution is a novel semantic model for encoding verbal arguments in multiple languages. The methodology is rooted in the k-Multilingual Concept (MCk) model, a state-of-the-art automated system designed for retrieving and aligning semantically-equivalent lexical items across k different languages. We integratedWordNet, BabelNet, and VerbNet into a framework that accommodates the unique demands of verbal context. The methodology is implemented in a highly-scalable pipeline, creating VerbAligNet, a new resource that encodes over 6k verbal arguments for 600+ verb senses, showcasing prevalent usage patterns across 9 valency frames on three languages. The evaluation demonstrates its accuracy in extracting semantically-equivalent verbal arguments for diverse verbs.
There is growing interest in "gaming the system" behavior, where students in online learning environments seek to progress without engaging in authentic learning processes. This study addresses this issue, a...
详细信息
ISBN:
(纸本)9783031643149;9783031643156
There is growing interest in "gaming the system" behavior, where students in online learning environments seek to progress without engaging in authentic learning processes. This study addresses this issue, aiming to automatically recognize this behavior in beginning programmers from a public and rural school in Northeast Brazil, as well as analyzing the demographic context of these students identified with this behavior. With the participation of 67 students, we collected data through student interactions with a programming learning environment, developing an automatic detection model. As a result, our detector based on the decision tree algorithm provided the best performance. Our findings highlight a significant difference between the group of students who exhibit "gaming the system" behavior and those who do not. Furthermore, younger students are more likely to exhibit such behavior.
In this paper, we propose to study the semantic development of the term immigrant in Croatian based on newspaper articles published in the online version of the Croatian newspaper Jutrarnji list from 30August 2015 to ...
详细信息
ISBN:
(数字)9783031566462
ISBN:
(纸本)9783031566455;9783031566462
In this paper, we propose to study the semantic development of the term immigrant in Croatian based on newspaper articles published in the online version of the Croatian newspaper Jutrarnji list from 30August 2015 to 17 January 2023. Our analysis is based on the approach of the Semantics of Argumentative Possibilities initiated by Galatanu [1], who stated that in the core meaning of every lexical term exists a possibility for deployment of meaning caused by the contextual environment. According to that, the core meaning of the word immigrant is calculated based on the lexical definition in the Croatian language. To test the environment of the keyword as the first step in constructing the network of contextual appearances of collocates and other appreciative elements, which can lead us towards the new semantic deployment of the key term, we developed a NooJ syntactic grammar. NooJ located occurrences of the key term and its partial synonyms, enabling us to select our key term's precise meaning. It also showed the immediate collocates of the term. We noticed that the grammar needed some improvement, considering the choice of prepositions in the syntactic grammar branches. However, that is operable, and can be used on larger corpora for further and more thorough analysis.
Since GPT-4's release it has shown novel abilities in a variety of domains. This paper explores the use of LLM-generated explanations as on-demand assistance for problems within the ASSISTments platform. In partic...
详细信息
ISBN:
(纸本)9783031643118;9783031643125
Since GPT-4's release it has shown novel abilities in a variety of domains. This paper explores the use of LLM-generated explanations as on-demand assistance for problems within the ASSISTments platform. In particular, we are studying whether GPT-generated explanations are better than nothing on problems that have no supports and whether GPT-generated explanations are as good as or better than teacher-authored explanations. This study contributes to existing literature since as of yet, there are no studies on the scale of ASSISTments evaluating the effectiveness of GPT support in education. Should GPT explanations prove effective then we plan to continue developing and evaluating explanations, hints, and other supports with GPT within ASSISTments.
This paper presents a systematic review of the scientific literature on trustworthy and ethical Artificial Intelligence (AI) and Education (AI&ED), including both AI applied in education to support teaching and le...
详细信息
ISBN:
(纸本)9783031643149;9783031643156
This paper presents a systematic review of the scientific literature on trustworthy and ethical Artificial Intelligence (AI) and Education (AI&ED), including both AI applied in education to support teaching and learning (AIED), as well as education about AI (AI literacy). Key interest is the identification of global trends with a special focus on unbalanced disparities. Strictly following the standardised protocol and the underlying PRISMA approach, 324 records were identified and selected according to the pre-defined protocol for the systematic review. Finally, 62 articles were included in the quantitative and qualitative analysis in response to four research questions: Which (i) journals, (ii) disciplines, and (iii) regions are leading scientific debates and sustainable developments in education and trustworthy/ethical AI, and (iv) what are the past trends? The articles revealed an unbalanced distribution across the various dimensions, together with an exponential growth over recent years. Building upon our analysis, we argue for an increase in interdisciplinary research that shifts the focus from the currently dominant technological focus towards a more human-centered (educational and societal) focus. Only through such a development AI can contribute effectively to the UN Sustainable Development Goal no. 4 of a world with equitable and universal access to quality education. The results of our systematic review provide the basis to address and facilitate equality in the future AI&ED progress across regions worldwide.
This article presents a sentiment analysis using data from X social media platform (Twitter) using artificial intelligence techniques. Two artificial intelligence techniques perform sentiment analysis: i) bag of words...
详细信息
ISBN:
(数字)9783031525179
ISBN:
(纸本)9783031525162;9783031525179
This article presents a sentiment analysis using data from X social media platform (Twitter) using artificial intelligence techniques. Two artificial intelligence techniques perform sentiment analysis: i) bag of words and ii) computer vision. The first is used for Natural Language Processing (NLP) and sentiment identification, while the second is for computer-based emotion identification in photographs or frames. The proposed methodology is applied to the soccer match between the Quer ' etaro White Roosters and the Atlas Football Club in Guadalajara, Mexico. The study case involves 2,000 tweets from the March 5, 2022, soccer match, collected from Twitter, and 200 photographs/images taken on the game day. The experimental analysis examined data by NLP in R language and computer vision using DeepFace. Results indicate negative sentiment perceptions with similar percentages of 74% for NLP and 81% for DeepFace, with an average negative perception of 77.5%.
FAIR Implementation Profile (FIP) is a special kind of linked data that consists of questions and answers about communities' decisions about the use of resources regarding the FAIR principles. Some FIPs have been ...
详细信息
ISBN:
(纸本)9783031659898;9783031659904
FAIR Implementation Profile (FIP) is a special kind of linked data that consists of questions and answers about communities' decisions about the use of resources regarding the FAIR principles. Some FIPs have been created to capture collective decisions by communities. However, FIPs have not been widely adopted by the communities in social science. In this paper, we explore how FIPs can capture the decisions of communities in social science by creating their FIPs and comparing them against existing attempts. The created FIPs could be used as a structured way of cataloging FAIR-related implementation, comparing the resources used across communities, as well as understanding the practice of FAIR principles. We perform our analysis by generating the FAIR Convergence Matrix and comparing the resources used and their use. Finally, we discuss the lessons learned and the limitations of this approach.
Recommender Systems in Virtual Learning Environments (VLEs) provide personalized suggestions to users based on preferences, interaction history, and behavior. They enhance learning by offering personalized content, in...
详细信息
ISBN:
(纸本)9783031643118;9783031643125
Recommender Systems in Virtual Learning Environments (VLEs) provide personalized suggestions to users based on preferences, interaction history, and behavior. They enhance learning by offering personalized content, increasing engagement, and improving teaching effectiveness. Challenges in VLEs include the cold start problem, data sparsity, and limited coverage. To address these, we propose G-Learn, a recommendation system operating in both supervised and unsupervised models. It utilizes graph machine learning, keyword mining, and similarity techniques to recommend educational materials tailored to each student's performance. We demonstrate G-Learn's effectiveness in a real scenario using data from Homero, a VLE for computerscience education developed for the brazilian federal government. Validation shows an average f1-score of 0.64 in unsupervised model and 0.95 in supervised model.
暂无评论