Research Of Enterprise Knowledge Management System Based On UIMA

As an evolution of information management, knowledge management is of great significance to promote the competitiveness of enterprises. Unstructured information, which is huge and grows rapidly, is the vital source of knowledge in enterprises. How to acquire, manage and use knowledge contained in the mass unstructured information is thereby an urgent problem to be solved. Although Unstructured Information Management Architecture (UIMA) can integrate a variety of Unstructured Information Management (UIM) techniques to annotate from unstructrured information, the annotations are difficult to be reused and shared for they are entity-centered but lack of relationship between entities and high coupled with applications.To address the problems above, this paper presents a Knowledge Acquisition Method based on UIMA (KAMU) which is based on analyzing the UIMA specifications and OWL standard. This method extends the basic type system defined in UIMA specification, and establishes the mapping between types defined in UIMA type system and the OWL classes and properties. Then, entity annotations are transformed to OWL individuals, and the relation annotations will be converted to OWL triples. KAMU proposes a domain ontology-based relationship extraction algorithm, which can extract relationships between entities with the relational model defined in ontology. In addition, KAMU supports detecting and eliminating semantic conflicts which exist in knowledge acquisition process through consistency checking. And implicit knowledge can be discovered through reasoning as well.Based on KAMU, this paper presents an UIMA-based solution for knowledge management system, then designes and implementes a prototype system. The system consists of three modules: unstructured knowledge source management, UIMA-based knowledge acquisition and reasoning, OWL knowledge-base management. The unstructured knowledge source management module use metadata to manage the heterogeneous, distributed knowledge sources uniformly; UIMA-based knowledge acquisition and reason module uses UIMA's API to integrate a variety of techniques to access and analysis information in knowledge sources and transform the result to the knowledge represented in OWL. The module also provide reasoning tools to support the consistency validation of the ontology and acquiring implicit knowledge; OWL knowledge-base management module provides management to OWL ontology such as version management, group management, distributed editing and so on. Under specific scenarios, the performance indicates that the system is feasible and effective.Contrasted with the traditional knowledge management system, this paper has the following characteristics:(1) An UIMA-based knowledge acquisition method called KAMU is presented. In use of the knowledge level type system KLTS, which is an extention of UIMA Base Type System, KAMU can transform the annotated result which are extracted from vast amounts of unstructured information by UIMA to knowledge represented in OWL. The method is scalable, flexible and adaptable;(2) An enterprise knowledge management system, EKMS, which is based on KAMU, is designed and implemented. With the combination of UIMA-based Information Extraction techniques and OWL-based Knowledge Management techniques, EKMS supports manage, analysis and knowledge acquiring from the distributed, heterogeneous and great amounts of unstructured information sources in enterprise. EKMS supports the distributed construction and editing of ontology as well.