A Web-scale Study of the Adoption and Evolution of the schema.org Vocabulary over Time

A Web-scale Study of the Adoption and Evolution of the schema.org Vocabulary

over TimeRobert Meusel, Christian Bizer and

Heiko Paulheim

A Web-scale Study of the Adoption and Evolution of the schema.org Vocabulary over Time - WIMS 2015 2

Motivation - LOD Cloud with 1.000 data providers


Motivation - schema.org MD with 700k data providers


Microdata in a Nutshell

- Adding structured information to web pages• By marking up contents and entities

- Arbitrary vocabularies are possible • Practically, only schema.org is deployed on a large scale

• Plus its historical predecessor: data-vocabulary.org

- Similar to RDFa

<div itemscope itemtype="http://schema.org/PostalAddress"> <span itemprop="name">Data and Web Science Group</span> <span itemprop="addressLocality">Mannheim</span>, <span itemprop="postalCode">68131</span> <span itemprop="addressCountry">Germany</span></div>


Schema.org in a Nutshell

- Vocabulary for marking up entities on web pages• 675 classes and 965 properties (as of May 2015, release 2.0)

- Promoted and consumes by major search engine companies• Google, Bing, Yahoo!, and Yandex

• Google Rich Snippets

- Community-driven evolution and development


Schema.org in a Nutshell – Coverage

- Schema.org has incorporated some popular vocabularies, like:• Good Relations (2012)

• W3C BibExtend (2014)

• MusicBrainz vocabulary (2015)

• Automotive Ontology (2015)


Microdata with Schema.org in HTML Pages

<html>…<body>…<div id="main-section" class="performance left" data-sku="M17242_580“>

<h1> Predator Instinct FG Fußballschuh </h1><div>

<meta content="EUR"><span data-sale-price="219.95">219,95</span>…</body></html>

HTML pages embed directly markup languages to annotate items using different vocabularies

<html>…<body>…<div id="main-section" class="performance left" data-sku="M17242_580" itemscope itemtype="http://schema.org/Product"><h1 itemprop="name"> Predator Instinct FG Fußballschuh </h1><div itemscope itemtype="http://schema.org/Offer" itemprop="offers"><meta itemprop="priceCurrency" content="EUR"><span itemprop="price" data-sale-price="219.95">219,95</span>…</body></html>

1._:node1 <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://schema.org/Product> .

2._:node1 <http://schema.org/Product/name> "Predator Instinct FG Fußballschuh"@de .

3._:node1 <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://schema.org/Offer> .

4._:node1 <http://schema.org/Offer/price> "219,95"@de .

5._:node1 <http://schema.org/Offer/priceCurrency> "EUR" .

6.…


Wrap-Up

- Semantic annotations are used by more and more websites

- Entities on websites become machine-readable and machine-understandable

- schema.org together with Microdata is a success story • Promoted by search engine companies

• Deployed by over 17% of all websites [1] (over 700k data providers)

- Usage is more compliant to the schema than e.g. LOD [2]

[1] http://webdatacommons.org/structureddata/2014-12/stats/stats.html[2] Meusel and Paulheim, ESWC 2015

http://webdatacommons.org/structureddata/2014-12/stats/stats.html



9

Digging for Reasons

- So, Microdata is more often deployed and is often more schema compliant, although there are millions of uncontrolled providers with different skill sets

- But why? Some hypotheses…• Availability of documentation

• Tool support

• Business incentive

• Schema flexibility

- Can we confirm/reject those from looking at the data?

A Web-scale Study of the Adoption and Evolution of the schema.org Vocabulary over Time - WIMS 2015

10

A Diachronic Perspective

- Versions of schema.org are archived over time• Plus: there are several crawl releases per year

• i.e., we can look at change over time

- If we look at both schema and deployed data, we may observe• Adoption rates of schema changes

• Data-first changes to the schema

• Convergence or divergence of deployed data


11

A Diachronic Perspective

- Three releases of WDC Microdata corpus [1]• 2012, 2013, and 2014

- Versions of schema.org that were valid• At the beginning of the crawl

• At the end of the crawl


[1] http://webdatacommons.org/structureddata




12

Top-down Adoption

- How fast are changes in the schema adopted?• New classes/properties

• Deprecations

• Domain/range changes

- Measuring adoption: challenges• Different crawls

• Overall growth of deployed schema.org

- Measure: normalized usage increase (nui) from i to j:• nui(s)>1.05: usage of schema element s has increased significantly

• nui(s)<0.95: usage of schema element s has decreased significantly


13

Top-down Adoption

- Adoption of new classes and properties• Almost half of all introduced classes are never used!

• Similar for new properties

- Reasons• Bulk-addition of vocabularies

• not every term is equally needed• e.g., medical vocabulary

• Blind spot of our approach• some terms are mainly for e-mail markup• e.g., Actions


SURPRISE!

14

Top-down Adoption

- Main domains of positive adoption• Meta data for web content

(schema.org/Website has the highest nui)

• Broadcasting (e.g., TV Episodes)

• Questions & Answers

• Postal addresses

- Classes featured in Google Rich Snippets• Still growth on high level (tens of thousands of data providers)

• But nui(s)<0.95


Yellow PagesSearch Engine Listings

Collaboration with BBC and EBU

Influence of CMS adoption

Q&A Pages, such asStackoverflow

15

Top-down Adoption

- Adoption of domain/range changes• Again: rather low overall adoption

- Adopted well for• Products (height, width, itemCondition, …)

• Broadcasting domain (episode, actor, ...)


Search Engine Listings

Collaboration with BBC and EBU

16

Top-down Adoption

- Adoption of deprecations• Works well (29 out of 32 have a significantly low nui)

- Exceptions• s:map (← s:hasMap)

• s:maps (← s:hasMap)


For Google Maps(lots of outdated tutorials)

17

Bottom-up Evolution

- Martin Luther• Started the protestant church

• A success story, too (like schema.org)

• (i.e., 800 million adopters worldwide)

- Famous quote:• “Man muss […] dem gemeinen Mann aufs Maul schauen”

• (roughly: “You have to listen to the way the common man really speaks.”)


Martin Luther, 1483-1546

Disclaimer:I do not speak for the

protestant church.

18

Bottom-up Evolution

- Are new features in the schema first used “inofficially”?• New classes/properties

• Domain/range changes

- Instrument for measurement: ROC curves• True positives mapped against false positives

• tp: elements used before

• fp: elements not used before

• Ranking by #PLDs


19

Bottom-up Evolution

- There are some mild influences observable• Stronger for domain/range changes

• especially range changes

• Weaker for new classes/properties


2012→ 2013 2013→ 2014 2012→ 2014

classes properties domains ranges

20

Bottom-up Evolution

- Extension mechanism• Allows for user-defined classes/properties

• Those become subclasses implicitly

- Analysis over time• No measurable impact on standard evolution

• “Inofficial” use is likelier than use of extension mechanism


s:Product/ElectronicProduct

s:price/reducedPrice

21

Overall Convergence

- Measuring convergence• i.e., homogeneity of descriptions of classes

• Example: two instances of s:LocalBusiness


_:1

_:2 “Birmingham”

“Main Street 24”

s:LocalBusiness

s:PostalAddress _:1

_:2 “Liverpool”

“Church Street 1”

s:LocalBusiness

s:PostalAddress

22

Overall Convergence

- Recap• RDF from Microdata is a set of trees

• i.e., we can enumerate all paths to leaf nodes(omitting literals)

- Example:


_:1

_:2 “Liverpool”

“Church Street 1”

s:LocalBusiness

s:PostalAddress

rdf:type-s:LocalBusiness, s:address-rdf:type-s:PostalAddress,s:address-s:addressLocality,s:address-s:streetAddress

23

Overall Convergence

- Using all paths, we can compute the entropy for each class as

- A low entropy refers to a high homogeneity

- We normalize both by maximum entropy and the total number of paths• i.e., we use normalized entropy rate as a measure for homogeneity


24

Overall Convergence

- Observations• Overall entropy decreases over time

- Classes with high convergence rates• WebSite, Blog, …

• Hotel, Restaurant, …

• Product, Offer, …

• Rating, Review


Influence of CMS adoption

Yellow pages

Google Rich Snippets

...all of the above

25

Key Adoption Drivers

- Search Engine Optimization• Web site providers want to be high in Google rankings

• Direct business incentive!

- Tool adoption• Major CMSs use schema.org

- Standard Agility• schema.org: 25 revisions in last three years

• cf. FOAF: six revisions in last eight years



Summary

- Both ways, top-down and bottom-up adoptions can be observed

- Homogeneity of deployed schema increase over time

- Described empirical data-driven study reveals valuable insights to understand how and why schema.org is a success story

- Observed key drivers and obstacles can also help to understand and analysis adoption of other standards, e.g. LOD

- More fine-grained insights might be revealed when extending the analysis corpus to the mailing list archive and issue tracker


Thank you! Questions? Feedback?

Raw data can be found on the website of WebDataCommons:

http://webdatacommons.org/structureddata/

More interesting datasets and analysis:

http://webdatacommons.org/index.html

Acknowledgement

The extraction and analysis of the datasets was supported by AWS in Education Grant.



http://webdatacommons.org/structureddata/index.html

http://webdatacommons.org/structureddata/index.html