A Causality Computation Retrieval Method with Context Dependent Dynamics and Causal-Route Search Functions

摘要

In this paper, we present causality computation methods and its application of a semantic associative search. We propose two essential methods for causality search, which are a causality computation method with context dependent dynamics and a causality route search method. The causality computation method with context dependent dynamics makes it possible to retrieve documents describing causal events in the context that specifies each situation of occurrent events. The causality route search method realizes to search respectively set of documents related to each generation of causal events from query events. We define three types of vector for each event data, that is, cause vector, effect vector and cause-effect vector that are characterized respectively with cause, effect and "cause and effect" event data. Applying a set of these vectors, our search method makes it possible to retrieve respectively "the document data describing cause events" and "the document data describing effect events" according to the context specified. Also, for realizing a causality route search, we construct query that represent sequential generations of causal events from a query event. Using the query constructed, we can retrieve documents about each generation of causal events. We have implemented a search system for an aerospace engineering field and clarified the effectiveness and the feasibility of our search method by several experiments.