Acquisition And Active Supply Of Experiential Engineering Knowledge Based On Context Awareness

The experiential knowledge(EK) in the brains of employees has been an important resource and a precious wealth of enterprise. With the sharp rise of employee mobility in recent years, enterprises urgently need to solve the accumulation and reuse problem of EK. However, as a kind of tacit knowledge, EK is difficult to be described and shared. To overcome the obscurity of EK definition and the inefficiency of EK acquisition and reuse currently, with the support of National Natural Science Foundation of China – project “evolvement principle and reuse method of empirical tacit knowledge supported by engineering semantics” and “active knowledge supply in engineering based on context awareness”, the doctoral dissertation studies the automated acquisition and active supply of EK in engineering domain.Based on classification of knowledge resources and comparison between knowledge attributes, the dissertation proposes the definition of engineering experiential knowledge(EEK). To do this a four dimentional knowledge attribute space is constructed for classifying knowledge according to its expressiveness, personalization degree, relationship uncertainty and context dependency, and the coverage of EEK is pointed out in the four dimentional space. Considering the core content of EEK, the dissertation proposes an ontology-based model to formally represent EEK and a nonmonotonic logic-based method to reason with EEK.The dissertation establishes a context model and the according context awareness mechanism oriented at computer-aided engineering(CAE), which are responsible for the timely capturing, filtering, associating and fusion of the information dynamically generated during the execution of engineering tasks. The research proposes four ways to acquire the associated context information, constructs functions and rules for measuring the influence of context information, and adopts a distance-based clustering method to fuse the context information. Through adjusting the weights of different distance items using supervised learning, the accuracy of separating blended task information is increased.The dissertation proposes an EEK acquisition method comprising experiential Q&A extraction, Q&A quality assessment and ontological concept mapping. Specifically, studies are carried out about Q&A pair recognition based on context awareness, Q&A association discovery and machine learning. Using the samples of high quality Q&A and low quality Q&A, the equations for computing Q&A quality are derived from logistic regression. The research builds EEK ontology semiautomatically from structured domain text, with which the acquisition of EEK is realized by mapping the Q&A sentences to the established EEK ontology. Semantic similarity calculation between words is used in the process of mapping sentences to ontology.The dissertation proposes an active EEK supply method based on context awareness. By dynamically constructing and matching knowledge application contexts duiring the execution of engineerin tasks, users’ knowledge needs are captured in real time and relevant knowledge is purposefully recommended to users. When matching the knowledge application contexts, a method combining statistical semantics and ontological semantics is proposed to measure the similarity between knowledge concepts. Regarding users’ current and future knowledge needs, a basic knowledge supply method and a proactive knowledge supply method are proposed respectively. At last, by mining the associations between EEK items, an EEK map is created to realize visualized representation and navigation of supplied knowledge.A prototypal EEK management system comprising the context awareness module, the EEK acquisition module and the EEK supply and navigation module is developed. With some typical tasks in the CAE domain as the application background, the proposed methods are applied and tested. Specific cases are used to show how the knowledge needs of engineers are perceived and satisfied by the prototypal EEK management system. By comparing with some alternative methods developed from state-of-the-art information retrieval techniques, the advantage of the proposed method in terms of knowledge supply accuracy and intelligence is clearly shown.