Insta-Search: Towards effective exploration of knowledge graphs

Abstract

Knowledge Graphs (KGs) are used to store heterogenous information in the form of graphs. One flexible and non-expert way to query these KGs is to use relationship queries or keyword search. The user can specify a query using keywords referring to entities in the graph. The system then returns a set of relationships among the queried entities. However, effectively querying these graphs is still challenging for a new user. She is not familiar with the entities and relationships in the graph and hence, her queries could often return empty or too few answers. We demonstrate a system called Insta-Search which facilitates effective exploration of KGs using relationship queries. Insta-Search helps the user by giving autocomplete keyword suggestions for partially typed words. It also displays an estimated number of answers that the current query would fetch along with few approximate top-scoring answers. The users also get entity suggestions so that they can iteratively reformulate the query until they find the query with the expected results. On submitting the query, the system returns ranked query results, grouped on the basis of similar information content to enhance result interpretation. No prerequisite knowledge of the data is required by the user to be able to use the system.