1-2hit |
Hirofumi KATSUNO Hideki ISOZAKI
Modeling a complicated system as a multi-agent system is one of the most promising ways of designing a large, complex system. If we can assume that each agent in a multi-agent system has mental states (beliefs, knowledge, desires and so on), we can formalize each agent's behaviors in an abstract way without being bothered by system implementation details. We present semantic structures that are useful for representing belief states in multi-agent environments. One of the structures is a restriction of partial Kripke structures studied by Jaspars and Thijsse: we assume that each agent can access from a state of a structure to at most one state. We call the restricted structures only-child partial Kripke structures. We show some properties of only-child partial Kripke structures. Another structure is a restriction of the alternate nonstandard structures defined by Fagin et al. to deal with the logical-omniscience problem. We show several relationships between partial Kripke structures and the restriction of alternate nonstandard structures. Using the results, we show that the outputs of a belief estimation algorithm we previously developed can be characterized by using only-child partial Kripke structures. Finally, we show that only-child partial Kripke structures are more appropriate for the belief estimation problem than the restricted nonstandard structures.
Tsutomu HIRAO Kazuhiro TAKEUCHI Hideki ISOZAKI Yutaka SASAKI Eisaku MAEDA
In this paper, we propose a machine learning-based method of multi-document summarization integrating sentence extraction with bunsetsu elimination. We employ Support Vector Machines for both of the modules used. To evaluate the effect of bunsetsu elimination, we participated in the multi-document summarization task at TSC-2 by the following two approaches: (1) sentence extraction only, and (2) sentence extraction + bunsetsu elimination. The results of subjective evaluation at TSC-2 show that both approaches are superior to the Lead-based method from the viewpoint of information coverage. In addition, we made extracts from given abstracts to quantitatively examine the effectiveness of bunsetsu elimination. The experimental results showed that our bunsetsu elimination makes summaries more informative. Moreover, we found that extraction based on SVMs trained by short extracts are better than the Lead-based method, but that SVMs trained by long extracts are not.