XPloreRank: exploring XML data via you may also like queries

Abstract

In many cases, users are not familiar with their exact information needs while searching complicated data sources. This lack of understanding may cause the users to feel dissatisfaction when the system retrieves insufficient results after they issue queries. However, using their original query results, we may recommend additional queries which are highly relevant to the original query. This paper presents XPloreRank to recommend top-l highly relevant keyword queries called "You May Also Like" (YMAL) queries to the users in XML keyword search. To generate such queries, we firstly analyze the original keyword query results content and construct a weighted co-occurring keyword graph. Then, we generate the YMAL queries by traversing the co-occurring keyword graph and rank them based on the following correlation aspects: (a) external correlation, which measures the similarity of the YMAL query to the original query and (b) internal correlation, which measures the capability of the YMAL query keywords in producing meaningful results with respect to the data source. Due to the complexity of generating YMAL queries, we propose a novel A* search-based technique to generate top-l YMAL queries efficiently. We also present a greedy-based approximation for it to improve the performance further. Extensive experiments verify the effectiveness and efficiency of our approach.

Knowledge Graphs: Opportunities and Challenges

“…Some recommender systems utilize the content of the original query result to discover highly-related items for the users that may interest them (Naseriparsa et al. 2019a ). These systems employ machine learning techniques or statistical measures such as correlation to compute the highly-similar items to those that are visited by the users (Naseriparsa et al.…”

XSnippets: Exploring semi-structured data via snippets

“…They focus on result diversification of keyword search results and exploiting the underlying XML data statistics to retrieve meaningful results for the user. In [32], a recommendation system called XPloreRank is proposed that uses two correlation scores to generate "You May Also Like" keyword queries for the users over XML data. However, XSnippets is the first interactive exploratory search framework that navigates a clueless user through her exploration over XML data by using XML snippets.…”