主催: The Japanese Society for Artificial Intelligence
会議名: 2012年度人工知能学会全国大会(第26回)
回次: 26
開催地: 山口県山口市 山口県教育会館等
開催日: 2012/06/12 - 2012/06/15
The internet revolution has brought about fundamentally new possibilities for people located at opposite sides of the globe to real-time dynamically communicate with each other. Although people most often use English as a common communication code, misunderstandings are almost unavoidable in cross-cultural communications.When focusing on problems at the lexical level, two questions might immediately arise: How can one improve the transfer of original conceptual meanings of source concepts to a target audience when there are no 100% equivalent concepts existing between the two cultures in consideration? And how does cultural prior-knowledge influence in such a communication scenario? The answers might be identified in Sperber & Wilson's Relevance Theory of Communication (1986) and the Knowledge Model involved in category-based inductions (Murphy, 2004). These cognitive theories imply that an audience's prior-knowledge is used for his/her inferential process of learning new information and such prior-knowledge is organized in a taxonomic hierarchy in his/her memory. Supported by these ground theories, a framework of bridging culturally-dependent ontologies by applying the Bayesian Model of Generalization (Tenenbaum & Griffiths 2001) is proposed. This approach enables one to estimate probabilities of how an information receiver generalizes a source concept from a given stimulus in a cross-lingual context.