Google Knowledge Graph is more complicated than public knowledge graphs when it comes to retrieving and reusing data for e.g. building custom KGs or concept maps. This is because even though there is a dedicated graph search API, it only o ers information about individual entities without links to other relevant resources that are only available in Google search knowledge panels. In this paper, we present Knowledge Graph Viewer, a tool that utilizes both the graph search API and knowledge panels in order to obtain relationships between graph entities to reconstruct custom substructures of the large knowledge base. This demo will showcase its usage and functionalities in each step of the KG creation while explaining the concept behind the work ow.
Google Knowledge Graph (GKG) is used by Google to store information about
entities and their relationships, which can then serve, among others, to enrich
the web search results. As with other large knowledge graphs (KGs), a question
arises if arbitrary fragments of it can be retrieved to create custom KGs for a
speci c use. Public KGs such as Wikidata or DBpedia1 o er exible open data
access via RDF dumps or SPARQL endpoints, on which tools such as concept
mappers2 can leverage. However, accessing graph-shaped information from GKG
is more di cult, as the GKG API only provides information about individual
entities (nodes) and not about relationships (edges). As observed by Barrasa [
Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). 1 https://www:wikidata:org/, https://dbpedia:org/ 2 E.g. ContextMinds { https://www:contextminds:com/, which dynamically autocompletes a manually built concept graph using data from DBpedia 3 https://www:blog:google/products/search/about-knowledge-graph-andknowledge-panels/ (e.g. birthplace, book author, works of a musician, categories, etc.), including the entities often sought together via Google search. An example search result with the knowledge panel can be seen in Fig. 1 on the right side. Using these pieces of information, we can progressively build meaningful graph fragments. However, manually copying relationship information from knowledge panels to connect nodes in a custom KG is slow and error-prone. To make this process easier, we developed the Knowledge Graph Viewer tool that can create custom KGs based on information from Google search knowledge panels and GKG API. 2 Knowledge Graph Viewer is a web application that uPsreofsilesJavaScript on both the back-end (in the form of NodeJS) and front-end side, Twwit ehrerInsetagratmheFacNebooukxtTu.mjbslr frYaouTmube ework (respectively Vue.js) is used, with data being stored in a Neo4j database. The application can: 1) add GKG entities, along with their relationships to a custom graph; 2) display the custom graph; and 3) export the graph and the underlying part of an RDFS ontology to a RDF, CSV or JSON le. taylor swift
Al Images Videos News About 517,000,000 results (0.78 seconds)
Shopping More Set ings Tools
Bilboard Taylor Swift's '1989' Now One of Only Four Albums by Women to Spend 300 Weeks on… 6 hours ago More for taylor swift
Bilboard Katy Perry Hopes Her Daughter Keeps Taylor Swift's Adorable Baby Gift Forever 1 day ago Kanye West Compares Music Biz and NBA to Slavery, Taylor Swift Roped In 8 hours ago www.taylorswift.com
The Of icial Website of Taylor Swift - folklore album out now! Event Details, Videos, Merchandise & More.
Taylor Swift Store · Events · The “in the trees” hoodie . . · So much for summer love . . ht ps:/ twit er.com/taylorswift13
Bout to show up at your party @ACMawards pic.twit er.com/ekNJfv4… Twit er · 2 days ago en.wikipedia.org › wiki › Taylor_Swift
Taylor Alison Swift (born December 13, 1989) is an American singer-songwriter. Her narrative songwriting, which often centers around her personal life, has received widespread media coverage. Born in West Reading, Pennsylvania, Swift relocated to Nashvil e, Tennessee in 2004 to pursue a career in country music.
LOathbeerlsn:aBmigeMs:aNchilsinSe;jöRbeeprgublic YGeeanrrsesa:cPtiovpe;: c2o0u0n3t–ryp;reroscekn;tsynth-pop; cou… Album · Folklore (Taylor Swift album) · Discography · 1989 (Taylor Swift album) More images
American singer taylorswift.com
YouTube Spotify Deezer
More music services Taylor Alison Swift is an American singer-songwriter. Her narrative songwriting, which often centers around her personal life, has received widespread media coverage. Born in West Reading, Pennsylvania, Swift relocated to Nashvil e, Tennessee in 2004 to pursue a career in country music. Wikipedia Born: December 13, 1989 (age 30 years), Reading, Pennsylvania, United States Height: 1.78 m Net worth: $360 mil ion (2019 estimate) Movies: Cats, Miss Americana, Valentine's Day, The Giver, MORE
View 35+ more
The source code is available at https://github:com/skaryys/kgviewer under the MIT license. A demo version is running at http://89:221:219:40/, and a video demonstration is available at https://vimeo:com/439597230.
In accordance with GKG usage terms4, we believe users can view the same data that is also publicly available from Google search results. Most of this data is obtained using Puppeteer5, a headless internet browser capable of retrieving data from web pages. Knowledge Graph Viewer should only be used for academic purposes. The tool has been tested in major web browsers, such as Chrome 29+, Firefox 77+, Opera 77+ and Edge 84+. 2.1
Adding GKG entities to the custom graph After entering a search term to the input eld, the result entities are displayed. Clicking on the Add to graph button then adds the entity into a queue, which can be seen on the left side of the page. Entities in the queue are then asynchronously added to our graph (see Fig. 2). A summary of this process is shown in Fig. 3.
If the entity in the knowledge panel does not match the entity obtained using the GKG API, or if there is no knowledge panel at all, only one node corresponding to the original entity is to be added to the custom graph. 4 https://developers:google:com/knowledge-graph/terms 5 https://pptr:dev/
Two ways of retrieving related entities from the knowledge panel are distinguished. For some links, the related entity ID can be obtained directly from the source code of the primary knowledge panel (see Fig. 4). However, for most links in the knowledge panels, the relevant entity IDs cannot be obtained directly from the source code of the original entity being acquired. For this reason, Puppeteer is used to open the search result page of the related entity. The process of adding a related entity is shown in Fig. 5.
From the overall procedures that our tool uses, it is evident that adding an entity to the custom graph along with related entities and relationships is a complex task that often requires opening several web pages. This is why some entities may take longer than others. In average, the whole process takes around one to two minutes. 2.2
Displaying the custom graph In Fig. 6, we show a custom graph built around the entity Prague and primarily containing information about important persons associated with the city. The graph visualization is based on Neo4j's native Neovis.js library6. By default, the tool shows up to 300 graph elements (nodes and relationships between them) to avoid slow loading, but this value can be adjusted manually. The resulting graph can also be searched and customized as follows: { Detailed information about entities is displayed on mouse hover. { Entities can be searched by name as well as by type. { Entities and relationships can be queried using the Cypher7 query language. { Users can specify which entity types should be displayed as additional nodes. 6 https://github:com/neo4j-contrib/neovis:js/ 7 https://neo4j:com/developer/cypher-query-language/ { Users can use `fast mode' which decreases rendering quality to improve loading speed. For graphs with many edges (e.g. if types are shown as nodes), this fast mode can reduce loading speed signi cantly.
{ Users can export results in RDF format. 3
Conclusions To our knowledge, Knowledge Graph Viewer is the only tool supporting the custom reconstruction of GKG fragments. Its usability is currently limited by two inherent factors: the slowness of entity addition due to multiple calls to the GKG API and Google search, and the necessity to adapt the wrapper interface when the structure of the Google search results changes. Nevertheless, we believe that it provides an added value compared to purely manual GKG reconstruction. As future work, we are preparing to explore the possibility of adding entities in an automated way, based on a user-set pro le. The tool itself can still be improved, i.e. not all related links must be browsed because there is direct connection between Google API identi ers and the stick parameter in Google search8.