Uncertain reasoning method and device for contexts

Abstract

The embodiment of the invention discloses an uncertain reasoning method and device for contexts. The method comprises following steps: calculating condition relevancy between context nodes of bottom layers, wherein condition relevancy represents condition dependency relationships between context nodes of bottom layers; setting up sides of context nodes of two bottom layers of each context node pair in N context node pairs; selecting a contact node in a bottom layer of N bottom-layer context nodes as a root node and pointing all set sides towards the direction which root nodes point to the rest nodes; setting directed edges from higher-layer nodes to context nodes of all bottom layers in order to set up a tree augmented nave bayes model; and utilizing the tree augmented nave bayes model to uncertain reasoning of contexts.The embodiment of the invention is capable of enhancing accuracy and efficiency of uncertain reasoning of contexts.

Description

Technical field [0001] The present invention relates to the field of communication, and in particular to a method and device for context uncertainty reasoning. Background technique [0002] Contextual reasoning is currently a widely used information reasoning method. Contextual reasoning often includes low-level context and high-level context. The low-level context includes information directly obtained by the device, and the high-level context includes information that the device cannot directly obtain. For example, in reasoning about the conference context, the high-level context may include the state of the conference room, and the low-level context may include the state of the computer in the conference room, the state of the projector in the conference room, the state of the microphone in the conference room, the noise level of the conference room, and the number of people in the conference room. In actual reasoning, the above-mentioned underlying context may be uncertain. F...

AI Technical Summary

Problems solved by technology

Among them, the underlying context node in the BN model is the parent node of the high-level context node, and there are too many directed edges between the underlying context nodes, and the increase of directed edges leads to the increase of the conditional probability in the BN model, and leads to the BN model's The complexity is very high, so when the sample data is small, the accuracy of the BN model reasoning is not high, and the reasoning efficiency is not high

Images