Study On Semantic Analysis And Formalization Of Product Classification

Product classification describes the properties of a product, determines its code structure, and simplifies the data of a product to make regular operation easier. Such classification establishes unified standard for the products of different organizations and producers, ensuring the real-time information of products to be systematic, identical, and sharable. In the age of big data, trades take place every second of every day, producing massive information and data associated with the merchandises. Generally, there are two ways to deal with massive data:one way is to find the connections between data based on Data Mining, which mines data that are not standardized, not structuralized, and half structuralized; the other way is to put emphasis on initial organization of information, a method that standardizes and structuralizes the data in the first place. Product classification is the application of the later way. However, product classification has five in-born disadvantages:the conflict between the linear and map structure of classes, the incapability of description, the limitation to the specificity of concepts, the extended update time unit and the high expertise required to operate on product classification.Semantic analysis is the premise to overcome the five in-born disadvantages. To make the utilization easier for users, preceding researchers have already devoted a lot of efforts, proposing the idea that use semantic approaches to overcome in-born disadvantages and even completely composing the eCl@ss with OWL Lite. Sadly, there is still lack of advanced semantic analysis on product classification itself, elimination of illogical semantics, and clarification on semantics. Therefore, to overcome the limitation of linear structure of classes, reinforce network connections, enrich semantics and specify the classes, we must conduct advanced semantic analysis on product classification.Formalization is the knowledge representation of product classification, which should meet three qualifications. First, formalization should thoroughly represent the system structure of product classification; secondly, formalization should have qualified semantic expressing capability to clearly express all kinds of semantic relations in product classification. Thirdly, the reasoning in formalization should be decidable, providing specific result in limited time.Concerning the systematic and practical problems of product classification, this thesis adapts manual semantic analysis to figure out the semantic relations within the classification, then formalizes the semantic relations with description logic, and transforms formal match into semantic match. With the ontology providing object relational mapping for the transitions between different classes of products, the search efficiency of product classification has been greatly improved.The ontology this thesis constructed is not meant to replace the current product classification, but meant to offer users access to product classification, which maps the users'inquiries on product to the product classification. As the connection between difference classification systems, the ontology, based on semantic match, offers translations between different systems.The major contents of this thesis can be summarized as follow.Chapter 1 IntroductionBased on the understanding of the meaning of product classification and the evaluation on its current status, this thesis proposes the research project targeting the disadvantages of product classification.Chapter 2 Product ClassificationBased on the origin and history of product classification, this thesis elaborates on the significance of product classification to our economy. According to four standards, scope of application, recognition, accuracy and accessibility, this thesis selects GPC and eCl@ss as the research object from six major international classification systems. By comprehensive introduction to GPC and eCl@ss and deep analysis on their system structures, this thesis describes the blueprint of ideal product classification.Chapter 3 the Formalization of Product ClassificationTransform product classification from offering information to offering knowledge requires the formalization of product classification. Among several formalization method, this thesis find the tool for knowledge presentation, description logic which is qualified for expression, and capable of, finite in and decidable in reasoning. By gradual introduction to description logic, this thesis avoids circular reasoning, proves the decidability of reasoning and reaches the conclusion that using description logic to replace nature language in defining concepts is impossible. Then, this thesis introduce the OWL language to achieve description logic.Chapter 4 Semantic Analysis and Ontology ConstructionSemantic Analysis ensure the veracity of concepts. This thesis analyzes the defects in specification of concept, adjustment of classes, operation on properties, and distinction on semantics. Then, this thesis proposes the principles to deal with these detects, handles them and constructs the ontology with the result of semantic analysis.Chapter 5 Ontology QueryThe query process matches the keywords with the ontology of product classification and reaches the query result by inference machine with certain rules and SparQL. By analyzing the query result, this thesis proves the veracity of the concepts, the completeness of semantics and the decidability of reasoning required to construct the ontology.Chapter 6 Conclusion and ProspectiveThis thesis answered how to solve the five in-born disadvantages and in what extant they are solved, and also indicates the solution, with in what extent it works, to the six difficulties in constructing and utilizing the ontology of product classification proposed by Mr. Martin Hepp. This thesis also mentioned the deficiencies and possible continuations of this research.The main contributions of this thesis can be listed as follow.? This thesis brings up four standards, scope of application, recognition, accuracy and accessibility, to select a specific proper semantic formalization and product classification from multiple choices. According to these four standards, this thesis selects GPC and eCl@ss as research objects.? By gradual introduction to description logic until SHOIN(D), this thesis avoids circular reasoning, and proves the theory, "if the definition of classes can be expressed by description logic, the machine will be able to determine what the classes are", to be incorrect.? By detailed semantic analysis, this thesis points out the defects in GPC and eCl@ss about specification of concept, adjustment of classes, operation on properties, improves these defects, distinguishes the semantics and constructs the ontology based on GPC and eCl@ss.? This thesis builds the test ontology query system, and, by adding reasoning rules into the system, proves the veracity of the concepts, the completeness of semantics and the decidability of reasoning.In response to five in-born disadvantages, this thesis makes certain accomplishments, which are described as follow.? The conflict between the linear and map structure of classesThe ontology of product classification takes the level system from its origin, but initiates networks between classes. Therefore, such ontology mediates this conflict by constructing the networks without damaging the linear structure.? The incapability of descriptionThe ontology of product classification distinguishes the relations between classes and emphasizes on the properties of classes. With the knowledge of product classification enriched by the compilation of property directories, the capability of semantic expression has been greatly improved.? The limitation to the specificity of conceptsThe ontology of product classification breaks the hierarchical limits of the original product classification, organizes the illogically arranged classes and, therefore, achieves the theoretical specificity of concepts.? The extended update time unitUnfortunately, the ontology of product classification is unable to solve this problem and there is no progress on the trend.? The high expertise required to operate on product classificationThe ontology of product classification expands the semantics on a massive scale. Therefore, the hierarchical, equivalent, opposite and class-property relations will greatly improve the accessibility for users.