Beschreibung

The predominant form of micro-content today is inspired by flashcards and are based on multiplechoice questions with additional elements, that provide guidance, feedback or further reading. In particular, in the context of this work, we will refer to the following five elements of a single microcontent unit: (a) question, (b) correct answer, (c) two to four distractors, (d) a hint, and (e) feedback and explanation. The objective of AMiGa is to deliver a complete automatic micro-content generation system. The system will receive as input digital learning material, together with a variety of options (e.g. preferred language, difficulty level, key phrases, etc.) and generate, as output, multiple-choice questions supported by adaptive feedback for learners. The multilingual QG/QA proposed models in literature are still limited and the zero-shot learning still suffers from the low quality of generated output. In AMiGa, we use a pretrained multilingual environment which enables using the system in different languages (see subsection 1.2.1.6). Our priority is first to support a selected set of languages through training data, and to further explore the possible quality improvements for zero-shot cross-lingual AQG scenarios. We extend the multilinguality support to the generation of distractors, hints and the adaptive feedback which introduces an additional value for our trained models that will be available to the research community.

Details

Projektzeitraum 01.01.2022 - 30.06.2024
Fördergeber FFG
Förderprogramm
Department

Department für Weiterbildungsforschung und Bildungstechnologien

Zentrum für angewandte Forschung und Innovation für lebensbegleitendes Lernen

Projekt­verantwortung (Universität für Weiterbildung Krems) Isabell Grundschober, BEd, BSc, MA
Zum Anfang der Seite