Research On Medical Ontology And Knowledge Acquisition

In AI, ontologies are scientific theories which can be used to interpret meta-properties of knowledge. Generally, an ontology is an explicit specification of a shared conceptualization, and it is fundamental in knowledge sharing in different agents and across different applications.The thesis aims to develop a medical ontology for knowledge acquisition. We have selected to use the term of domain-specific ontologies (DSO), and attempted to design an applicable Medical Domain-Specific Ontology (MDSO). Upon MDSO, we have established a subset of NKI, called NKIMed. During the research on NKIMed, we have considered the medical concept space (MCS) design and medical ontology (MDO) design, and exploitedmethods of automatically extracting medical knowledge from semi-structured text.We have acquired 52 medical categories, 1691 attributes, 107 relationships, 554 items of clinical test knowledge, 19595 items of medical concepts, and totally 78012 pieces of medical knowledge. Meanwhile, we have also designed numerous medical axioms for checking the consistency of the NKIMed, knowledge inference and interconnection between different medical concepts. We have designed algorithms for checking the consistency of knowledges and axioms in NKIMed.To tackle the problem of knowledge acquisition, we have introduced Knowledge Acquisition Agents (KAA) and Knowledge Acquisition Pattern (KAP) and summarized a number of Medical KAA (MKAA) and Medical KAP (MKAP). To make MKAA work, we have designed a context-sensitive grammar and relevant algoritms. To resolve conflicts in multi-agents, the Relative Knowledge Match Degree (RKMD) is computed. We have designed a semantic guess algorithm to work with RKMD to clear away conflicts in multi-agents.Two applications of NKIMed, i.e. intelligent teaching systems and speech diagnosis are illustrated.