Rather than generating a finished diagram, the system suggests keywords and propositions a teacher can accept, edit, or ignore.
The Kit-Build method asks students to rebuild a concept map from a set of ready made pieces. It works well in classrooms, but it leans on one heavy prerequisite. A teacher has to prepare a solid concept map for every reading before the lesson can happen.
That preparation is slow. Doing it well for a set of English reading texts can take hours per text, and teachers who are short on time simply skip the method. Automatic generation looks like an obvious fix, until you read the output. Maps produced automatically tend to come out too general for a specific lesson.
The researchers took the middle path. Their web based authoring tool mines a reading text and offers suggestions instead of finished diagrams. A teacher selects a sentence, reviews what the system proposes for it, and places the useful parts onto their own canvas.
Three expert teachers tested the tool across 15 English reading texts, each around 719 words and 34 sentences long. For every text the system suggested roughly 157 keywords, 85 verb based propositions, and 62 open domain propositions. The suggestions arrive grouped in three sections inside a support panel.
The accuracy picture was mixed, and the authors report it plainly. Keyword suggestions had high recall at 0.953 but low precision at 0.405 when borderline suggestions were judged strictly. Proposition triples scored lower on both counts.
Pinandito, A., Prasetya, D.D., Hayashi, Y., et al. (2021). Semi-Automatic Concept Map Generation Approach of Web-Based Kit-Build Concept Map Authoring Tool. International Journal of Interactive Mobile Technologies.
