Embodied Conversational Agents (ECAs) are interactive characters that exhibit human-like qualities, such as facial expressions, lip-synch, or emotional voice, and are able to communicate with humans, or with other ECAs by using natural human capabilities (speech, gestures, etc.). However, to make current ECAs' dialogue management strategies more appealing and real to the user, they should be aware of general knowledge about the external world. This factual knowledge, which is independent of personal experience, should be stored in their semantic memory. This paper presents a knowledge-based solution to improve learning through ECAs with factual knowledge based on semantics. In particular, we build this semantic memory by means of a novel proposal known as Daira. Moreover, we integrated Daira with Maxine, a powerful animation engine for developing applications with embodied animated agents. To illustrate the potential of our approach, we designed a proof of concept in which our system is able to provide data from the online Great Aragonese Encyclopedia (GEA), written in Spanish, to engage students. The experiments performed show the feasibility and efficiency of our proposal. In particular, we demonstrated that using enriched ECAS when searching information can enhance learning motivation and learning performance, making the interaction process much more accurate, simpler, and near to the students.
