Information Technology Intelligent Search of Content in E-commerce Systems

Abstract

The article describes the process of developing intelligent search technology for content for the implementation of the module of e-commerce systems for forming a list of recommendations for regular users. Intelligent search of content is based on methods of linguistic analysis, modern algorithms for parsing and finding words, and recommendations based on user preferences. The main components of such a search are the parsing of text strings, the selection of keywords, the spelling check, the recognition of common abbreviations and acronyms, the semantic analysis of the text, the search by relevance with the extraction of synonyms, filters and sorting. A web application based on Java and Elasticsearch was developed with the implementation of a recommender system based on a collaborative filtering algorithm. The purpose of the work is to develop the technology of intelligent product search with the formation of a list of recommendations for the user. The object of the research is the processes of intelligent search with the possibility of generating recommendations for users in the field of any e-commerce without reference to the categorization of goods/services, etc. The subject of research is the methods and means of intelligent search of recommender systems based on the Collaborative Filtering algorithm for the formation of product recommendations for users, which is oriented on general coincidences of the choices of similar users. During the experimental testing of the developed system, a number of search queries were conducted with and without the NLP algorithm, the results of which demonstrated an improvement in system performance within the range of 15–95 % depending on the keyword and the presence/absence of errors in the search words. A comparison of the speed of execution of requests with already existing systems was also carried out. Yes, the amount of data in the storage may differ (error when comparing 60–70 ms). For example, a query that consists of 1 or 2 words will be found much faster by 20–70 ms compared to its counterparts. But for 3 and more, results are about the same – 9–20 ms faster.