=Paper=
{{Paper
|id=Vol-2980/paper307
|storemode=property
|title=InK Browser -- The
Interactive Knowledge Browser
|pdfUrl=https://ceur-ws.org/Vol-2980/paper307.pdf
|volume=Vol-2980
|authors=Joseph Zalweski, Lu Zhou, Cogan Shimizu, Pascal Hitzler
|dblpUrl=https://dblp.org/rec/conf/semweb/ZalweskiZSH21
}}
==InK Browser -- The
Interactive Knowledge Browser==
https://ceur-ws.org/Vol-2980/paper307.pdf
InK Browser – The Interactive Knowledge
Browser
Joseph Zalewski, Lu Zhou, Cogan Shimizu , and Pascal Hitzler
Data Semantics Lab, Kansas State University, USA
{jzalewski, luzhou, coganmshimizu, hitzler}@ksu.edu
Abstract. We present an improved implementation of the Interactive
Knowledge Browser (InK Browser), a tool for exploring knowledge graphs
visually, using a schema diagram.
1 Motivation
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 [5],
graphical representations of an ontology, to provide this context when navigating
the graph. These diagrams can be ingested quickly, typically having few enough
elements that they can be fully displayed on a single page, and show the large-
scale structure of the data in the graph. Of course some ontologies are vast, and a
specimen like the Gene ontology, with thousands of classes, would not be a good
choice of ontology for this purpose. Yet many useful ontologies applicable to large
KGs are small, and the InK Browser exploits these kind of ontologies and their
schema diagrams. In theory there is no limit to the size of the knowledge graph
that can be explored this way, since the tool is built on the existing SPARQL
technology, which shows good scalability to large graphs.
2 Related Work and Contributions
Visual knowledge graph browsers already exist, some at an enterprise deploy-
ment level and in practical use [1][2]. However, they mostly focus on displaying
0
Copyright © 2020 for this paper by its authors. Use permitted under Creative
Commons License Attribution 4.0 International (CC BY 4.0)
1
See http://wifo5-03.informatik.uni-mannheim.de/pubby/.
2 Zalewski, J., Zhou, L., Shimizu, C., Hitzler, P.
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
differs 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 fit 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 [3]. In particular, the current version is implemented in an ex-
tensible object-oriented style, and supports access to graphs hosted on remote
endpoints, as well as control of the layout of diagrams.
3 Implementation Details
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 SPARQL-
Wrapper Python library. The browser consists of a collection of views. Views
communicate with each other using a simple, flexible message protocol in order
to propogate user actions across multiple views. This extensible plug-in philoso-
phy 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 file, such as the annotations generated while using CoModIDE [6].
This schema diagram in this view is interactive and will allow a class-level
navigation by clicking on the nodes of the graph.
– Class Hierarchy: This view displays the class hierarchy of the KG in a
collapsible tree format, as mapped to a common schema, shown in Figure
1b in order to provide users with an overview of general topics in the KG.
Child-parent relationships indicate subsumption. Currently, we display this
mapping according to the Wikipedia Category Graph [4]. Mappings are gen-
erated via a pre-trained word2vec model to detect the most similar term in
2
See jgraph.github.io/mxgraph/ and https://jgraph.github.io/mxgraph/
3
See http://daselab.org/content/modular-ontology-engineering-portal.
InK Browser – The Interactive Knowledge Browser 3
the Wikipedia category graph. We plan to support Schema.org4 or SUMO
ontology5 as well.
– Entity Annotation: This view displays annotation assertions for a se-
lected entity and is shown in Figure 1b. Currently we automatically dis-
play rdfs:label, rdfs:comment, and dcterms:provenance, but hope to make this
customizable in the future.
– Text Search: The text search view provides users with a search box in the
current workspace. The search view takes the user’s input and returns a list
of similar entities with their IRIs and useful annotations sorted by relevance.
This can help users target their understanding of the KGs. An example is
shown in the Figure 1d.
– Graph Statistics: The graph statistics view displays simple graph statistic
(e.g., numbers of nodes, edges, and triples) and is shown in Figure 1e.
4 Demonstration
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 [6].
5 Conclusion and Future Work
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 finally, 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.
References
1. Gruff – new browser based version. https://allegrograph.com/products/gruff/,
accessed: 2021-08-14
4
See https://schema.org/.
5
See https://www.ontologyportal.org/.
6
See https://knowwheregraph.org/.
4 Zalewski, J., Zhou, L., Shimizu, C., Hitzler, P.
(a) A look at the InK Browser. The left column displays distinct instances of the class
chosen in the schema diagram (right column) and specific instance data for a given
choice of individual (center column).
(b) Wildfire Class in Hierarchy (c) Wildfire Annotations
(d) Text Search View (e) Graph Statistics View
Fig. 1: Text Search and Graph Statistics View
InK Browser – The Interactive Knowledge Browser 5
2. Knowledge graph visual browser web application. https://github.com/
linkedpipes/knowledge-graph-browser-frontend, accessed: 2021-08-14
3. Chittella, R.S.: Leveraging schema information for improved knowledge
graph navigation (2019), http://rave.ohiolink.edu/etdc/view?acc_num=
wright1564755327091243
4. Sarker, M.K., Schwartz, J., Hitzler, P., Zhou, L., Nadella, S., Minnery, B.S., Juvina,
I., Raymer, M.L., Aue, W.R.: Wikipedia knowledge graph for explainable AI. In:
Villazón-Terrazas, B., Ortiz-Rodrı́guez, F., Tiwari, S.M., Shandilya, S.K. (eds.)
Knowledge Graphs and Semantic Web - Second Iberoamerican Conference and
First Indo-American Conference, KGSWC 2020, Mérida, Mexico, November 26-
27, 2020, Proceedings. Communications in Computer and Information Science,
vol. 1232, pp. 72–87. Springer (2020). https://doi.org/10.1007/978-3-030-65384-2 6,
https://doi.org/10.1007/978-3-030-65384-2_6
5. Shimizu, C., Eberhart, A., Karima, N., Hirt, Q., Krisnadi, A., Hitzler, P.: A
method for automatically generating schema diagrams for modular ontologies. In:
1st Iberoamerican Conference on Knowledge Graphs and the Semantic Web (2019),
to Appear.
6. Shimizu, C., Hammar, K., Hitzler, P.: Modular ontology modeling. Tech. rep. (2021),
https://daselab.cs.ksu.edu/publications/modular-ontology-modeling, un-
der Review.