Towards a conversational agent architecture to favor knowledge discovery in serious games

  • Authors:
  • Francesco Bellotti;Riccardo Berta;Alessandro De Gloria;Elisa Lavagnino

  • Affiliations:
  • University of Genoa, Italy, Genova;University of Genoa, Italy, Genova;University of Genoa, Italy, Genova;University of Genoa, Italy, Genova

  • Venue:
  • Proceedings of the 8th International Conference on Advances in Computer Entertainment Technology
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

"Conversational Agents" (CAs) are virtual characters controlled by the computer and able to dialogue with users in natural language. CAs are usually employed in virtual world applications, such as for training, gaming or advertising, in order to increase the situation realism and user involvement. We are implementing a natural language interaction system with a clear focus on instructional dialogues aimed at favoring a player's acquisition of knowledge on specific topics through interaction with specialized NPC. Another key requirement for the system is easy and efficient writing and maintenance of the texts, also by author with no specific expertise on computational language technology. The system we propose includes a strategy and a tactical level. The former is responsible for managing the high-level aspects of the conversation, while the latter responds to the player's queries by relying on an original combination of an essential syntactical analysis and an ITF-IDF-based procedure. The technique specifically targets a (educational) gaming application domain. However, we believe that the technique could be extended to meet the requirements of other application domains.