Study On Product Family Designing And Processing Methods For Online Customer Reviews

With the development of Internet and the deepening application of Web2.0 technology, more and more customers post their opinions concerning products, services, brands or corporations on the traditional internet and mobile internet, these online reviews or opinions have been greatly influencing customers'purchasing decisions and firms' reputations and performances. Online Customer Reviews (OCR) contains a huge commercial value, however, how to polymerized use it is a big challenge. Web2.0 technology solves the original and scale problems of OCR, but the "disordering, decentralization" OCRs bring a great deal of information fragments, which has increased information overload and information lost for customers in the Internet world. In other words, OCR will become tininess, trivial and fuzzy-valued if it cannot be ordered and centralized. For the Internet enterprises, OCR is a kind of product showing information or knowledge, and its added value comes from producing, processing and using it. Therefore, how to transform the "disordering, decentralization". OCR into the "ordering and centralization" OCR in order to meet different customer demand, has become an outstanding problem both in the academic field and business practice, and it is worthy of further study. Under the above-mentioned background, this dissertation aims to help Internet enterprises with OCR providing and takes the information fragments solving as the breakthrough point. The author adopts several theories and methodologies such as knowledge management, production and operation management, product family design theory, schema, and hypergraph to conduct a systematic and further study from the following aspects.First, this dissertation proposes a total processing model directing at the core processing task of "ordering and centralization" OCR. In particular,1) To define OCR and analysis its processing characteristics; 2) To define what's the "order and ordering" and "center and centralization" of OCR; 3) Based on several relative theories, this study puts forward a three-dimension model from the dimensions of demanding, ordering and centralization and makes a formal analysis. In addition, an actual example is given to test the model's feasibility based on the improvement of processing way of OCR from existing e-commerce sites.Second, the product family design of OCR processing includs the entire processing process, and it can be classified into conceptual model domain from the software engineering angle. Specifically,1) To establish a hyper-graph design model of OCR product family designing based on the adoption of axiomatic design theory and hypergraph theory, so that the complex product family designing process can be transformed into a kind of formalized description; 2) To give a new mapping method according to the function and physical domain of hyper-graph design model, and demonstrate the mapping process between and within different domains. Furthermore, an algorithm is presented to evaluate and optimize the mapping result, and finally the mapping method proves to be feasible through an instance.Third, presenting the OCR scheme that is a knowledge expression of production processing way and a guide for main processing through presenting prior knowledge, and it can be classified into logical model domain from the software engineering angle. In this dissertation, three production processing schemas with different functions and added values are designed based on the sensorimotor schema theory, which are:1) Pretreatment schemas containing data cleaning, data conversion and metadata extraction; 2) Ordering processing schemes including object extraction, polarity classification and value analysis, etc.; 3) Centralizing processing schemes expressing entity, event and activity process. This research explores on the structures of the above three types of schemes with specific problems of input, output and transformation, and the matching problem among different processing schemes.Forth, exploring the hyper-graph system of OCR which is the supportive method of production processing realization. Modular primitives are the simple sub-patterns at the bottom which are discussed in this study from the perspective of hyper-graph system, and it can be classified into physical model domain from the software engineering angle. Specifically, to establish hypergraph system models presenting the schemes of ordering and centralizing processing, and to explore the expression of data relationships. After that, Modular primitives are extracted and their evaluations are further studied.In the end, under an application background of OCR in food area, this dissertation analysis the opinion leader identification of OCR, and establishes the pretreatment, ordering and centralization schemas, and then to present the actual expressions of data relationships. After that, a comparison is given between the opinion leader identification method of this study and that of a commercial site (dianping.com). Finally, the value and advantage of OCR processing method in this study is tested.