Data Wrangling: Generating a U.S. Presidential Election Knowledge Graph from a Google Spreadsheet

Arising from curiosity about the potential outcome of the 2020 US Presidential Election, I’ve constructed a spreadsheet comprising data about eligible voters in said election, a breakdown of early-votes cast by state, and the final tallies for the 2016 election.

I am interested in states that show very high early-vote counts as a percentage of the total vote count in the prior 2016 Presidential Election.

Data Wrangling Goal?

I want to generate an explorable Knowledge Graph from my Google Spreadsheet which itself comprises a merge of data from various sources that include:

• DBpedia
• 2016 Election Dataset published to Github by Kris Shaffer
• 2020 Early Voting Data from the U.S Elections Project

I will use the Virtuoso instance behind the public URIBurner Service (which includes an enabled Sponger Transformation Middleware Module), as my tool of choice. 

Steps?

[1] Create Google Spreadsheet. Here’s my example. 

[2] Pass the Spreadsheet URL to URIBurner using options for returning CSV rather than HTML — https://linkeddata.uriburner.com/about/html/https/docs.google.com/spreadsheets/d/1PTqUkqv-9BPWY1V1cFq13xR92EL6vJ9b8N4Ox3TUagE/gviz/tq?tqx=out:csv&range=A2:O53&sheet=2020_Election_Analysis

That returns an HTML-Entity Description Page which is also an entry point into a newly generated Knowledge Graph. This pages includes a listing of Entities derived from Rows in the original Spreadsheet. 

For instance, clicking on the hyperlink “Record2” results in a lookup that returns data for the state of Texas.

[3] Faceted Browsing View — https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Flinkeddata.uriburner.com%2Fabout%2Fid%2Fentity%2Fhttps%2Fdocs.google.com%2Fspreadsheets%2Fd%2F1PTqUkqv-9BPWY1V1cFq13xR92EL6vJ9b8N4Ox3TUagE%2Fgviz%2Ftq%3Ftqx%3Dout%3Acsv%26range%3DA2%3AO53%26sheet%3D2020_Election_Analysis&distinct=1

Alternatively, you can click on the “Browse using” drop-down or manually construct an alternative URL-pattern to obtain a different HTML-based Entity Description page that includes powerful Faceted Browsing capability i.e., the use of Entity Attributes to provide pivot-style exploration across various dimensions.

For instance, here is a list of instances of the Class "2020 Election Analysis" derived from the Google Spreadsheet.

Here's a view of a specific Record, a specific instance of the "2020 Election Analysis" Class derived from the Google Spreadsheet.

Here's the Ontology derived from the Google Spreadsheet.

A Little Reasoning & Inference

At this point, I have two views of my Knowledge Graph, but the labelling for each state could be clearer. 

To solve the preferred labelling problem, I’ve applied a SPARQL INSERT statement to generate an additional attribute that will ultimately be used as my preferred label, courtesy of the skos:prefLabel term from the SKOS Ontology.

PREFIX lod:                 <https://lod.openlinksw.com/>
PREFIX rdfs:                <https://www.w3.org/2000/01/rdf-schema#>
PREFIX schema:              <https://schema.org/>
PREFIX owl:                 <https://www.w3.org/2002/07/owl#>
PREFIX skos:                <https://www.w3.org/2004/02/skos/core#>
PREFIX election-2020-data:  <https://docs.google.com/spreadsheets/d/1PTqUkqv-9BPWY1V1cFq13xR92EL6vJ9b8N4Ox3TUagE/gviz/tq?tqx=out:csv&range=A2:O53&sheet=2020_Election_Analysis#>
INSERT {
         GRAPH <urn:us:election:2020:data:cleanup>
           { ?record skos:prefLabel ?label }
       }
WHERE { ?record election-2020-data:State_Name ?label }

The following screenshots depict effects of the INSERT statement above i.e., how the underlying Faceted Browsing Engine has selected the preferred Entity Attribute for Display Labeling used in the “About:” heading. 

And here's a revamped view of "Texas" rather than "Record2" .

Conclusion

Courtesy of the data management and transformation power of Virtuoso, I’ve successfully used a single mouse-click to generate a Knowledge Graph deployed using Linked Data principles from my Google Spreadsheet — comprising data about the U.S. Presidential Election collated from a variety of sources.

Related

• Data Wrangling Definition from the OpenLink Technology Glossary 
• US 2020 Presidential Elections Google Spreadsheet
• Ontology Generated from the Google Spreadsheet -- RDF Turtle Format
• Ontology Generated from the Google Spreadsheet -- JSON-LD Format
• What is the Virtuoso Sponger, and Why is it Important? 
• Virtuoso Home Page