Video Search Engine Design Based On Text Retrieval

With the boom of Internet technology development,the online video industry has also risen and prospered.How to provide users with high-performance,high-availability video retrieval services,and enable users to efficiently and quickly retrieve the video information they desired has become the core topic of the video field research.To achieve the goal,this paper designs a video search engine based on video meta data retrieval.The bottom framework of the video search engine is based on the high-performance micro-services framework gRPC,which uses remote procedure calls to communicate between services and uses the interface definition language Protocol Buffers to generate service interfaces.As a natural distributed storage and searching server,elasticsearch,in terms of data storage,can be used to provide high-performance as well as high-quality data services.As for the business logic,this paper makes a video retrieval workflow and divides the business module according to the actual video search business requirements,and due to the completion of two tasks above,the service framework of the video search engine is designed.It also shows how this process is carried out from the user's query word into the video information.The process of generating the client and server under the gRPC framework is also explained.For intention recognition module,the core business logic of video search engine,this paper has proposed different recognition strategies to address different users' searching needs.For ordinary-videos and series-videos,the indexing rules are adopted to make the intention judgment of the corresponding query words,as for tags-videos,the tags recognition strategy based on AC automaton(Aho-Chorasick string match algorithm)word segmentation is adopted.When dealing with the meta data,this paper outlines the establishment of indexing rules to service intention recognition module and expounds its role in video retrieval services.In order to improve the user's product experience and take into account the different levels of users' needs,the collection of tag-videos searching results is divided into three dimensions of "hot","praise" and "latest" and two pages for good display.When the video search engine designed as well as implemented,it shows excellent performance.At the same time,troubleshooting strategies are proposed based on issues arising during its running time and technical points,which can be upgraded,are briefly showed.The purpose of what have done about troubleshooting strategies and upgraded points is to take fully account of the robustness and scalability requirements of a commercially valuable software.