A Vector Space Retrieval Method with Causal Relationship Computation Functions for Event Data

Abstract

In this paper, we propose a vector space retrieval method with causal relationship computation functions for event data. The method is used for the domain in which a query given by a user can be regarded as an event. By expressing an event as a vector, causal relationships between events can be described as numerical values. This feature enables to set a level of importance on a specific event flexibly. The method can be applied to document retrieval with causal relationship computation. We clarify effectiveness of the method by showing several experimental results using synthetic event and document data sets.