Abstract:

The national OER repositories are of limited use due to the language barrier. Slovenia has created advanced interactive i-textbooks and a collection of Khan Academy-style video lectures. We will translate these Slovenian materials into English as a pivot language using advanced machine translation models. Educational video lectures (EdVL) must follow several design principles. We analysed the literature and established several empirically verified characteristics of EdVLs. We classified them into four pedagogical and six technical principles. Some features of EdVLs are more pedagogical in nature, emphasising the importance of interactivity, generative activity, dynamic drawing, and the omission of seductive details. Technical principles emphasise recording perspective, teacher visibility, emotional state, gaze guidance, captions, and live composite recording format. In order to machine translate EdVLs, designers must follow an additional eleventh principle, the principle of machine translation. The machine translation principle enables successful machine translation of EdVLs. This principle includes several features that are important for designers who want to reach an international audience. Designers of such video lectures recorded in a non-English language must consider the possibility of omitting unneeded, language-impaired text. The lecturer must speak in clear, short sentences, avoid dialect, and deliberately convey emotion through grammar (verbal punctuation) and word choice. We chose STEM as the test domain and started with mathematics. Since mathematics is a highly symbolic language, video lectures that follow all eleven principles are machine translated without compromising the effectiveness of video lectures that target conceptual understanding of mathematics. We present the results of the ongoing development. The final results will be transferable to other school subjects (especially hierarchical subjects such as languages) and systems (e.g., the neural model of machine transcription and translation). Initial results in classroom implementation were obtained using qualitative and quantitative methods from social science education research. The evaluation took place in primary and secondary schools in Slovenia, Germany, Turkey, Finland and Spain. Three approaches were used for the evaluation: a) flipped classroom with intermediate level interactive EdVL (i.e. with prompt questions), b) flipped classroom with high level interactive EdVL (embedded in LMS), and c) EdVL with guided viewing during synchronous teaching. In a flipped classroom, the individual learning phase takes place before the group phase. Students are first given a learning resource (usually a video) to watch independently, and then build higher taxonomic knowledge in a group activity in the classroom. By adapting this teaching and learning approach, we expect to have a positive impact on deeper understanding and knowledge in STEM, effective use of technology in education, and encouraging policy makers to achieve higher levels of digital literacy among all stakeholders in the school system.

Keywords:

Video pedagogy, educational video lectures design, automatic transcription and translation, multilingual educational materials, supervised machine translation.