Design And Implementation Of The Location-based Keyword Query Recommendation System

Keyword query recommendation is a fundamental method to improve the quality and accuracy of the searching results.It plays a significant role in commercial search engines,enterprise database queries,e-commerce,and personalized information services.Its goal is to recommend alternative keyword queries to the users to retrieve desired information.With the rapid spread of smart phones,people have moved from desktop computers to mobile devices to search information in their daily life.Location-based search results not only require relevance to the searching keywords submitted by the user,but also to the user's current location.This is also referred to as the spatial keyword search.Existing keyword recommendation methods do not consider the location of the users and the spatial dimension of the information.However,many applications require that the retrieved information should be close to the user's current location,such as location-based services.This thesis designs and implements a location-based keyword query recommendation system.It adopts a recommendation model that considers both the semantic relevance of the suggested queries w.r.t.the user submitted queries and the spatial proximity of the results retrieved by the suggested queries based on the user's current location.The main contributions of the thesis include:1.Design a location-based keyword query recommendation system.The system adopts a client/server architecture and consists of three modules: a data module,a server module,and a client module.The data module collects and cleans the log data in the search engine and constructs a bipartite graph on the queries and the documents.The server module adjusts the model by incorporating the spatial proximity of the documents and uses the BCA algorithm to calculate the similarity between the submitted query and other queries.The client module displays the recommended keyword queries and the final search results in the context of the map.2.Implement the location-based keyword query recommendation system.The data crawled from www.dianping.com is used for the demonstration of the system.The dishes are considered as keyword queries and the restaurants are regarded as documents.A bipartite graph on the dishes and the restaurants is constructed.The user submits a dish with the current location.The server finds several dishes that are served in the nearby restaurants using the location-based keyword query recommendation model.3.A limitation of the system is that a new query which does not exist in the model cannot receive suggestions.The New Query Association Algorithm is proposed to solve this problem.Specifically,all the queries in the model are classified using an online classifier.When a new query comes,it is firstly classified using the classifier,and then the similarities between the new query and the queries in the same category are calculated.The top-k queries are selected and passed to the recommendation model.