Literaturnachweis - Detailanzeige
Autor/inn/en | Wu, Zhongdi; Larson, Eric; Sano, Makoto; Baker, Doris; Gage, Nathan; Kamata, Akihito |
---|---|
Titel | Towards Scalable Vocabulary Acquisition Assessment with BERT [Konferenzbericht] Paper presented at ACM Conference on Learning @ Scale (10th, Copenhagen, Denmark, Jul 20-22, 2023). |
Quelle | (2023), (6 Seiten) |
Zusatzinformation | ORCID (Wu, Zhongdi) ORCID (Larson, Eric) ORCID (Sano, Makoto) ORCID (Baker, Doris) ORCID (Gage, Nathan) ORCID (Kamata, Akihito) Weitere Informationen |
Sprache | englisch |
Dokumenttyp | gedruckt; Monographie |
Schlagwörter | Prediction; Vocabulary Development; English (Second Language); Second Language Learning; Second Language Instruction; Teaching Methods; Transfer of Training; Grade 2; Elementary School Students; Scoring; Artificial Intelligence; Learning Analytics; Networks; Reliability; Evaluation Methods; Science Instruction; Social Studies; Speech Communication; Task Analysis; Natural Language Processing; Computer Software Vorhersage; Wortschatzarbeit; English as second language; English; Second Language; Englisch als Zweitsprache; Zweitsprachenerwerb; Fremdsprachenunterricht; Teaching method; Lehrmethode; Unterrichtsmethode; Training; Transfer; Ausbildung; School year 02; 2. Schuljahr; Schuljahr 02; Bewertung; Künstliche Intelligenz; Reliabilität; Teaching of science; Science education; Natural sciences Lessons; Naturwissenschaftlicher Unterricht; Gemeinschaftskunde; Aufgabenanalyse; Natürliche Sprache |
Abstract | In this investigation we propose new machine learning methods for automated scoring models that predict the vocabulary acquisition in science and social studies of second grade English language learners, based upon free-form spoken responses. We evaluate performance on an existing dataset and use transfer learning from a large pre-trained language model, reporting the influence of various objective function designs and the input-convex network design. In particular, we find that combining objective functions with varying properties, such as distance among scores, greatly improves the model reliability compared to human raters. Our models extend the current state of the art performance for assessing word definition tasks and sentence usage tasks in science and social studies, achieving excellent quadratic weighted kappa scores compared with human raters. However, human-human agreement still surpasses model-human agreement, leaving room for future improvement. Even so, our work highlights the scalability of automated vocabulary assessment of free-form spoken language tasks in early grades. [This paper was published in: "Proceedings of the Tenth ACM Conference on Learning @ Scale (L@S '23), July 20-22, 2023, Copenhagen, Denmark," ACM, 2023.] (As Provided). |
Erfasst von | ERIC (Education Resources Information Center), Washington, DC |
Update | 2024/1/01 |