Research On The Key Techniques Of Product Information Retrieval In E-commerce

Electronic-Commerce has been developing with an astonishingly high pace since the beginning of the 21st century. The fact that online-shoppers in China have reached as many as 142 million is showing a bright future of the Electronic-Commerce market. With the Electronic-Commerce applications of middle and small sized enterprises getting common, online retailing becoming routines, online-shopping participants being improved both in quantity and in quality, online-shopping has become a significant element of business activities. The number of Electronic-Commerce sites is increasing and the resources of business information are accumulated, featured as diverse, extreme and heterogeneous. Information seems overloaded while the intended information users can access shrinks comparatively and the’information loss’phenomenon is common in the process of online shopping. It is hard for public search engines (Google and Baidu, etc.) to meet the users in recall ratio and product information in users’real need can’t be retrieved or gained conveniently and rapidly, making considerable time wasted on excluding information which is unrelated, repeated, redundant or undependable. Therefore, how to search information and products in users’real need swiftly and efficiently in face of such mass Electronic-Commerce data is the problem to be solved urgently in commerce information retrieval.Electronic-Commerce search engines have become a basic means for people to get products, service and information in Electronic-Commerce activities, an important element for the development of Electronic-Commerce and a hot spot where consumers, business and Electronic-Commerce researchers pay much attention. However, present Electronic-Commerce searching is faced with such handicaps:the single retrieval mode that makes Electronic-Commerce sites isolated; Electronic-Commerce information in the field binding problem; the satisfactory of users’personal and diverse need; the identification and process of the credibility of product information. These has brought new challenges to information retrieval and Electronic-Commerce.The research on the product information retrieval in Electronic-Commerce of this paper is a meaningful exploration conforming to the circumstance and responding to the requirement described above. The content of this paper is structured as follows.Part oneConstruct a retrieval service mode (ProductRank) of integral product information that contains the collection, organization, matching and recommendation of product information. At present, product information index services provided by most Electronic-Commerce websites basically apply traditional product retrieval mode focusing on keywords, so the retrieval system can’t recognize the users’real needs. In this paper, the author studies the complexity and extremeness of the mass data on the web and analyzes the present situation of web data crisis and its cause from the angle of the development environment of Electronic-Commerce as well as business and users’need; identifies the information feature of Electronic-Commerce sites and comprehensively researches the information index needs of Electronic-Commerce products; studies the present Electronic-Commerce search engines and the platform construction technologies and compares the engines home and abroad; discusses the characteristics of information process in Electronic-Commerce, chooses from the filtered products information, gets customers’and products’context then match them, at last discovers the feature and quality of the product on the basis of intelligently mining the previous buyers’comments, thus forming a complete product retrieval service platform that can collect, retrieve, organize product information and feedback the outcome by the credit level.Part twoPropose a product semantic crawler of product information based on Electronic-Commerce to realize the collection of product data. With the accumulation of web information, networks have become the hugest knowledge pool to us, where there is a lot of Electronic-Commerce information. However, none of the existed search engines such as Google, Baidu and etc. is intended for the information retrieval in a particular field, because they are open to different users, trying to return with the best results that takes everything into consideration. As a result, the data in a user’s real need is often drowned in mass of useless information. In the application of Electronic-Commerce, the trade information determines the retrieval range of vertical search engines so structuring an effective crawler is the key point to complete the collection and index of product information. In this paper, the crawler continuously optimizes the reference in Electronic-Commerce by studying ontology in its crawling process and combines the analysis of web topic links and semantic analysis of the content of web topic, taking advantages of which to reach a better field constraint effect, prevent the topic drift problems and bring in a higher topic harvest rate.Part threeBuild a user profile and product context model and put forward an algorithm that matching users with products relying on the calculation of context likeness, the ContextRank. In product information retrieval, there is a coexistence of personality and diversity of users’ needs, so the same to products. The author applies both users’need and product properties to the same model, tempting to imitate the context of users’need and product properties and interpreter users’requirement thoroughly. Moreover, the author tries a method of context perception, proposes a way of matching products with users’need by context calculation and gets a more precise rank of products according to the distance of semantic context, with the products that satisfy the user better getting higher ranks.Part FourCreate a credit recommendation method, the Opinion Rank, by mining the comments of the products. Faced with numberless products from various shops, customers get confused when they go shopping, especially when they are searching one kind of products from several choices of shops. And to some degree, sellers’boastful promotions of their products enhance the risk and difficulty of online shopping because it is hard for consumers to tell the right from wrong under the always positive influence of the advertising. After conscientiously analyzing the credit problems of product information, the author judges the credit of a product by the comments from buyers who have already purchased and used it, analyzes the properties and the positive or negative extreme degree of the properties by natural language translation, probability and statistics method and the machine learning theory, to precisely rank the products and make the product information retrieval system with enough credit to be recommended, thus returning with the wanted and reliable products to users.Part fiveDevelop a retrieval system of books named the BookRank on the theory above. The system catches web pages of books by the semantic focused crawler of products and forms an vertical retrieval space for books sold online; Perceive searchers’different context and matches the right book products with the certain context; Gathers the comments on the books and bring about a rank of recommended books by adding the extreme degree of the comments and allocating certain credit to each book accordingly. What’s more, an experiment of PER(Purchase expectation Ratio) is done to prove the retrieval quality and the author verifies the product effect of retrieval outcome ranking of the BookRank method via contrasting the experimental data, which emphasizes the importance of a powerful search engine within an Electronic-Commerce website. The comparison of experimental data explains the theory proposed in this paper to provide the consumers with valuable product information and worthy of application, making the system an experimental platform for further research, Which also reveals significance of the intense need for a Electronic-Commerce search engine.Generally speaking, this essay studies the means and model of cross-platform product information retrieval system to offer both the enterprises and customers access to product information of high effect and quality. It enables the sellers to organize product information and the buyers to get it in Electronic-Commerce application successfully, drafts a model for the technology compatibility of enterprise Electronic-Commerce platforms, optimizes the service outcome of product retrieval, filters useless information, matches the users’ context with the product semantically, explores the properties of products band on the intelligible mining of the former buyers’ comments and constructs a complete service centre for the collecting, retrieving, organizing and dependable recommending of product retrieval, thus making much contribution to the users’ gradually growing online shopping need. In terms of theory, this paper analyzes the characteristics of web extreme data, finds out the cause of mass of information loss, discusses the properties of Electronic-Commerce information and forms a set of methods in information index service. It contributes to the further development and expansion of the theory of information resource management in the information environment and the positive interaction of information theory with Electronic-Commerce activities by researching the problems in the information facet. In addition, the progress of applications and the proposal of new theories in Electronic-Commerce will bring in excitement even new development the inteligence information field.The content of this research is the achievements of the MOE Project of Key Research Institute of Humanities & Social Science at Universities:’The research on Electronic-Commerce information retrieval and restructuring by cross-platform’(No.07JJD870220) and Humanities & Social Science Research Project of Hubei Provincial Department of Education:’The research on the Electronic-Commerce information resource semantic management in Web data crisis’(No.2009b228).