14, NOVEMBER 2017 Andrea Volpini Co-founder & CEO at WordLift AI-Powered SEO Build Your Own Knowledge Graph and make your content easy to be found.
14, NOVEMBER 2017
Andrea VolpiniCo-founder & CEO at WordLift
AI-Powered SEO
Build Your Own Knowledge Graph and make your content easy to be found.
I help publishers expand traffic with
...using semantic web technologies
an organic sustainable growth
semantic web
@cyberandywww.smxl.it #SMXLmilan
Keep readers engaged by helping machines understand what your content is about
+18.47% New Users from Google
+17.3% Avg.
Time Spent
+13.75% Session Duration
after 3 months on a design blog from Poland
readers engaged
@cyberandywww.smxl.it #SMXLmilan
! download the research paper "
@cyberandywww.smxl.it #SMXLmilan
@cyberandywww.smxl.it #SMXLmilan
How many of you are using semantic markup? Insights from our favorite crawler
+
@cyberandywww.smxl.it #SMXLmilan
How long would it take to mention all relevant concepts in an article using Schema? relevant conceptsSchema
@cyberandywww.smxl.it #SMXLmilan
What is Linguistic Semantics?A preliminary introduction
PERSON
THING (REFERENT)
STRING (SYMBOL)
@cyberandywww.smxl.it #SMXLmilan
Sharing Meaning Across the Globe Here Comes the Semantic Web
Martha
Andy
birthPlace
Italy
Website A Website B Website C
knows
Linked Data
@cyberandywww.smxl.it #SMXLmilan
Computers use URLs as unique identifiers
ANDY ENTERS LINKED DATA
http://mywebsite.com/Andy
Andy presents himself to computers using a Linked Data persistent URL
unique identifiers
@cyberandywww.smxl.it #SMXLmilan
RDF in a nutshell
1. Assign URIs to entities (“things”) ○ Example: Andy the Person as the URI
ns:Andy 2. Link entities using typed
links (triples) ○ Example: ns:Andy, type, Person
3. An RDF model is a set of triples in a machine readable format
4. Publish on the web
ns:Andy
<schema:Person>
<Entiy>
“Andrea Volpini”
subClassof
rdf:type
rdfs:label
RDF
@cyberandywww.smxl.it #SMXLmilan
5-stars linked data markup5 stars linked data
@cyberandywww.smxl.it #SMXLmilan
While you write, WordLift…
☝Runs a full text analysis a. WordLift analyses the content of an article and
identifies matching entities and topics from large big data graphs such as Dbpedia, Freebase and Geonames or from the user’s internal vocabulary
b. All extracted concepts are organized in four categories: Who (Person, Organization), Where (Place), When (Event), What (Creative Work, Product, Intangible)
✌Transforms articles into an 5-stars linked data graph that machines can process % Provides meaningful recommendations to let readers explore your content
@cyberandywww.smxl.it #SMXLmilan
How 5-stars Linked Data affects findability Search is Everywhere
“Findability” is not simply SEO
It’s the ability to find and use content by establishing a context, by making content easier to be found and by guiding the user to the “next steps”
Building conversational interfaces that plug into Google Assistant, Cortana and Amazon Echo is part
of this process
@cyberandywww.smxl.it #SMXLmilan
Ok Google, what is personal assistant search
optimization ?
personalassistantsearch
optimization
@cyberandywww.smxl.it #SMXLmilan
“Findability” in the era of Semantic Search
Answer
#2
2 links instead of 1
Interpretation
Search Intent
@cyberandywww.smxl.it #SMXLmilan
Input Features
@cyberandywww.smxl.it #SMXLmilan
How can a computer understand human language?
Intent Knowledge Answer
“When is SMXL Milan 2017
taking place?”
“type”: “event”
“date-time”: “startDate”
…
Interpretation
Google gets the answer from the graph-based representation of knowledge (subject-predicate-object describing the event) that are found in our
webpage subject-predicate-object
@cyberandywww.smxl.it #SMXLmilan
Where does this answer come from?
author
Entity
Schema.Org - Make Your Data Shine
AUTHORSHIP MATTERS
URL
Article Website
@cyberandywww.smxl.it #SMXLmilan
The will be shown as a backlink of Boris’s website in the Google Search consol !
Article
Schema.Org - Let’s look at the data
Google Search Console
@cyberandywww.smxl.it #SMXLmilan
Moving from building pages to creating structured data helps you create relationships between entities that matters. Connecting all information related to an event like SMXL Milan across multiple pages has been strategic to stand out on Google search.
● Increased dwell time ● Enriched user experience with improved navigation ● New ways to engage readers with conversational interfaces
RELATIONSHIPS REALLY MATTER
@cyberandywww.smxl.it #SMXLmilan
The Entity Based Content Model
Ok Google, what is an entity in WordLift?entityWordLift
@cyberandywww.smxl.it #SMXLmilan
Let’s have a look at how content is organized
SMXL Milan 2017
<Event>
Andrea Volpini <Person>
Richard Wallis <Person>
Martha van
Berkel <Person
>
URL
image
type
relation
relation
relati
on
Metadata Pillar Content Cluster Content
@cyberandywww.smxl.it #SMXLmilan
… Wait, is this content or data?
It’s both!
<Person>
<Person>
<Person>
<Event>
Domain Specific Knowledge (Domain Entities and Properties) Meta Knowledge (Classes/Concepts)
subclassOfsubclassOf
subclassOf
subcla
ssOf<Place>
<Entity>
subcla
ssOf
@cyberandywww.smxl.it #SMXLmilan
How to make your website talk Long Live AI - The Web is dying and bots will be the new Websites
WordLift: all you need to knowChatbots
intent
USERKNOWLEDGE (WEBSITE &NLP & BOTMESSAGING
answer
intent + context query
resultsresponse
PLATFORM FRAMEWORK DATA GRAPH)
@cyberandywww.smxl.it #SMXLmilan
Chatbots: a simple use-case
“When is SMXL 2017 taking place?”
USERKNOWLEDGE (WEBSITE & DATA GRAPH)
NLP & BOT FRAMEWORK
MESSAGING
PLATFORM
“The event will start November the 13th”
@fact-category = webinar @sys.date-time = startDate
Surface = Phone
curl 'https://wordlift.io/blog/en/wp-
admin/admin-ajax.php' \ --data 'action=wl_jsonld&id=5758'
"startDate":"2017-11-13"
Text response + “startDate”
We use the same on-page structured data that Google uses on its SERP
Chatbots
@cyberandywww.smxl.it #SMXLmilan
@fact-category
Interpretation
SPARQL Query
Result
SELECT * WHERE { BIND( <http://schema.org/Event> as ?type )
?subject a ?type ; rdfs:label ?label ;
schema:startDate ?startDate . FILTER ( xsd:dateTime( ?startDate ) > now() )}
We use a query language called SPARQL to compute the results
“When are the next events ?”
Chatbots: a more advanced way of pulling data from your graphChatbots
events@fact-category
@cyberandywww.smxl.it #SMXLmilan
1. By entity >> (show me all content on AI) 2. By date >> (show me all contents on AI written
in the last two weeks) 3. By engagement >> (show me all contents on AI
that have more than 20 interactions) 4. By author >> (show me last 10 articles by Ben
Dickso n on AI) 5. By type >> (show me all articles that have to
do with our events) 6. By class >> (show me all articles that have to
do with dbo:Software)
WordLift Knowledge Graph - Querying Capabilities
AIAI
last two weeksAI
Ben Dickson AI
events
dbo:Software
@cyberandywww.smxl.it #SMXLmilan
Get smarter with your own data Semantics on Analytics: let’s turn entities into insights
<EVENT SPARQL QUERY GOES HERE>
Sessions by EntitiesSessions Entities
@cyberandywww.smxl.it #SMXLmilan
When does harvest (ernte in German) peak?
@cyberandywww.smxl.it #SMXLmilan
harvest
Questions?
@cyberandywww.smxl.it #SMXLmilan
@cyberandywww.smxl.it #SMXLmilan
Hey Google, talk to
Doctor Search Marketing
Andrea Volpini
https://wordlift.io
https://github.com/insideout10/wordlift-plugin This work is licensed under a Creative Commons
Attribution 4.0 International License.
@cyberandywww.smxl.it #SMXLmilan
“I don't believe you”, is the first thing I said to the Wordlift team when they told me what their plugin could do. For us, as a publisher of business articles and a total dedication to original content, the benefits seemed almost too good to be true.
Creating good content is hard enough in itself. Plugins like Wordlift make us feel like that effort isn’t in vain and our stories will be found by the right people, bringing them inspiration and knowledge.
Your Content deserves this.Ramia Marielle El Agamy, Editor-in-Chief at Tharawat Magazine