Search Box System to Improve User Interaction in Knowledge Graph Searches: A Solution for Users Without Expert Skills

Abstract

Knowledge graphs enable semantic search in a variety of information systems. Some collaborative and public ones have existed for more than a decade. That is the case of Wikidata, one of the most popular knowledge bases today. It contains a graph of entities that has been growing since 2012 to host more than 94 million items. Currently, the exploration of data in Wikidata is a tedious task, especially for inexperienced users: it requires users to go deep into entities and perform a large number of clicks and searches of property names. Our goal is to allow nonexpert users who want to enter new data in Wikidata to be able to examine its ontologies in an agile way so that they can know whether a particular statement is present in the graph. To this end, we propose a new search box for Wikidata, which allows an exploratory search based on chained entities without having to navigate through pages. It is based on the real-time classification of the information that the user enters in the search box and the dynamic generation of suggestions based on the exploration of entities and relationships. A quantitative analysis is performed comparing how many iterations a user needs to perform a set of popular searches with our proposal and standard Wikidata search. The results suggest that interaction can be reduced by about 43%.