We present an improved implementation of the Interactive Knowledge Browser (InK Browser), a tool for exploring knowledge graphs visually, using a schema diagram.
There are many ways to explore data in a knowledge graph (KG). One can
run SPARQL queries, for instance. Thus, a set of specialized tools have been
developed to streamline the process, such as knowledge graph browsers (e.g.,
Pubby1). These provide a user experience similar to browsing the Web itself:
entity URIs become hyperlinks, and triples involving that entity are presented,
with further clickable entities, allowing the user to traverse the graph, while
seeing only a small, focused piece of it at a time. In some cases, we have found
that such browsers may be too locally focused: they do not allow the user to orient
themselves within the graph as a whole. We propose to use schema diagrams [
Visual knowledge graph browsers already exist, some at an enterprise
deployment level and in practical use [
Commons License Attribution 4.0 International (CC BY 4.0)
1 See http://wifo5-03.informatik.uni-mannheim.de/pubby/.
arbitrary nodes of a knowledge graph and therefore need to be highly dynamic,
since it is usually impossible to meaningfully show all nodes at once. InK Browser
di ers from these products in that it does not display most entities visually, but
only ontological concepts, thus avoiding the explosive complexity of visuals that
often results from visual browsing, since a small static set of concepts can be
used that t on the screen all at once. The whole browsing experience in InK
Browser is run through this static diagram. This work is a reimplementation
and extension of [
The project is built using Flask (which powers the browsing functionality) and
mxGraph (which renders our schema diagram).2 Actions within the browser
are tied to generated SPARQL queries, which are executed using the
SPARQLWrapper Python library. The browser consists of a collection of views. Views
communicate with each other using a simple, exible message protocol in order
to propogate user actions across multiple views. This extensible plug-in
philosophy makes it easy to customize functionality, and thus the user's perspectives on
the graph. We list the currently available modules below. Implementation and
installation details for this tool and tool suite can be found in our online portal.3
{ Instance List: This view displays a subset of instance data for a selected
class and is shown in Figure 1a (left column). The class shown is determined
when the user clicks on a new entity, such as an entry in the Instance Data
view or node in the Schema Diagram view.
{ Instance Data: This view displays the \triple" information of a selected
entity in a traditional KG browser style, shown in Figure 1a (middle column).
{ Schema Diagram: This view displays the schema diagram for the KG, as
shown in Figure 1a (right column). This prototype uses an external script
to extract the relevant schema diagrammatic information from a prepared
OWL le, such as the annotations generated while using CoModIDE [
We will demonstrate the InK Browser live. The demo will include accessing a
real KG developed as part of the KnowWhereGraph6 project, and navigating
it using the GUI functionality of InK Browser. The schema diagram used to
organize the data will be one created from CoModIDE [
InK Browser shows promise as a new paradigm in knowledge graph browsing, allowing the user to see individual triples and the structure of an entire graph at once. However there is much to do moving forward. We anticipate to improve the UI of the program (e.g., improving its visual look and feel), the UX (e.g., identifying an rdfs:label or concept name to use instead of an IRI), and the overall integration of Flask's features, develop additional modules (e.g., a map-based geographic data interface that detects and displays available spatially-explicit data), and nally, conduct a user study to more objectively determine whether the InK Browser browsing experience is any better than that of existing systems. Acknowledgement. The authors acknowledge support by the National Science Foundation under Grant 2033521 A1: KnowWhereGraph: Enriching and Linking Cross-Domain Knowledge Graphs using Spatially-Explicit AI Technologies. 4 See https://schema.org/. 5 See https://www.ontologyportal.org/. 6 See https://knowwheregraph.org/. (a) A look at the InK Browser. The left column displays distinct instances of the class chosen in the schema diagram (right column) and speci c instance data for a given choice of individual (center column).
(b) Wild re Class in Hierarchy (c) Wild re Annotations (d) Text Search View (e) Graph Statistics View Fig. 1: Text Search and Graph Statistics View