Presented at the Scottish Linked Data Interest Group workshop
https://www.eventbrite.co.uk/e/3rd-scottish-linked-data-interest-group-workshop-registration-9568925915
Abstract:
Know-how available on the Web, such as step-by-step instructions, is largely unstructured and isolated from other sources of online knowledge. To overcome these limitations, we propose extending to procedural knowledge the benefits that Linked Data has already brought to representing and retrieving declarative knowledge. We describe a framework for representing generic know-how as Linked Data and for automatically acquiring this representation from existing resources on the Web. This system also allows the automatic generation of links between different know-how resources, and between those resources and other online knowledge bases, such as DBpedia. We discuss the results of applying this framework to a real-world scenario and we show how it outperforms existing community-based integration efforts.
More than Just Lines on a Map: Best Practices for U.S Bike Routes
Integrating Know-How in the Linked Data Cloud
1. 3rd Scottish Linked Data Interest Group workshop
Linking Data…
what about Know-How?
Paolo Pareti
PhD student at Edinburgh University
10 September 2014
9. What is Know-How in this Context?
Focus on Human Know-How:
● instructions created by humans for humans
● tasks where humans are (so far) the main agents
10. Some Important Properties
of Human Know-How
Human Know-how can be:
● vague (different interpretations)
● erroneous
● missing (common sense is often assumed)
● distributed (related knowledge in different repositories)
● diverse (multiple approaches, possibly contradicting)
● evolving (knowledge changes, new knowledge is created)
11. Some Important Properties
of Human Know-How
Human Know-how can be:
● vague (different interpretations)
● erroneous
● missing (common sense is often assumed)
● distributed (related knowledge in different repositories)
● diverse (multiple approaches, possibly contradicting)
● evolving (knowledge changes, new knowledge is created)
13. Why Human Know-How as Linked Data?
machine
understandable
integration with
other knowledge
reusable
open data
better search
and discovery
common sense
task reasoning
activity
recognition
automation
software/robotic
immediate
benefits
future
benefits
data
knowledge
14. Why Human Know-How as Linked Data?
machine
understandable
integration with
other knowledge
reusable
open data
better search
and discovery
common sense
task reasoning
activity
recognition
automation
software/robotic
immediate
benefits
future
benefits
data
knowledge
16. How can Know-How be Represented?
How to notify a client that the
requested order has been dispatched?
17. How can it be represented?
Procedure to notify a client that the
order has been dispatched:
1. Make a copy of the dispatch
receipt
2. Send the receipt to the client
3. Send a confirmation email to the
client's email address
18. How can it be represented?
Procedure to notify a client that the
order has been dispatched:
1. Make a copy of the dispatch
receipt
2. Send the receipt to the client
3. Send a confirmation email to the
client's email address
Linked Data
P0
19. How can it be represented?
Procedure to notify a client that the
order has been dispatched:
1. Make a copy of the dispatch
receipt
2. Send the receipt to the client
3. Send a confirmation email to the
client's email address
Linked Data
P0
P1 P2 P3
20. How can it be represented?
Procedure to notify a client that the
order has been dispatched:
1. Make a copy of the dispatch
receipt
2. Send the receipt to the client
3. Send a confirmation email to the
client's email address
Linked Data
P0
P1 P2 P3
?
21. How can it be represented?
Procedure to notify a client that the order
has been dispatched:
1. Make a copy of the dispatch receipt
2. Send the receipt to the client
3. Send a confirmation email to the
client's email address
Linked Data
P0
P1 P2 P3 How to send a receipt to the client:
1. Put the receipt in an A6 envelope
2. Write the client's delivery address
3. Send using a 1rst Class signed
stamp
S0
S1 S2 S3
?
22. How can it be represented?
Procedure to notify a client that the order
has been dispatched:
1. Make a copy of the dispatch receipt
2. Send the receipt to the client
3. Send a confirmation email to the
client's email address
Linked Data
P0
P1 P2 P3 How to send a receipt to the client:
1. Put the receipt in an A6 envelope
2. Write the client's delivery address
3. Send using a 1rst Class signed
stamp
S0
S1 S2 S3
24. How can it be represented?
Procedure to notify a client that the order
has been dispatched:
1. Make a copy of the dispatch receipt
2. Send the receipt to the client
3. Send a confirmation email to the
client's email address
How to send a receipt to the client:
1. Put the receipt in an A6 envelope
2. Write the client's delivery address
3. Send using a 1rst Class signed
stamp
How to notify a client that the
order has been dispatched:
1. Make a copy of the dispatch
receipt
2. Send the receipt to the client
3. Send a confirmation email to
the client's email address
25. How can it be represented?
Procedure to notify a client that the order
has been dispatched:
1. Make a copy of the dispatch receipt
2. Send the receipt to the client
3. Send a confirmation email to the
client's email address
How to send a receipt to the client:
1. Put the receipt in an A6 envelope
2. Write the client's delivery address
3. Send using a 1rst Class signed
stamp
How to notify a client that the
order has been dispatched:
1. Make a copy of the dispatch
receipt
2. Send the receipt to the client
a. Put the receipt in an A6
envelope
b. Write the client's
delivery address
c. Send using a 1rst
Class signed stamp
3. Send a confirmation email to
the client's email address
27. Experiments and Results
Is it possible to extract the Linked data representation of
procedures and then link it with other related knowledge?
1. Knowledge Extraction
2. Knowledge Integration
28. Knowledge Extraction
Automatic knowledge extraction from semi-structured
websites
Results (from crawling the WikiHow and Snapguide websites):
● 211,696 extracted procedures
● 2,609,236 extracted entities
29. Knowledge Integration
Two kinds of integration of know-how:
● with existing Linked Data
● with other know-how
30. Integration with existing Linked Data
Extracted
Know-How
Dataset
How to make a pancake
requires
2 eggs
http://de.dbpedia.org/page/Egg
is a
31. Integration with other Know-How
Extracted
Know-How
Dataset
How to make a pancake
has step
break the eggs
has method
How to break an egg
33. Know-How Integration:
Results
Community Integration* Automatic Integration
Number of links 106,056 221,351
Precision 70.7% 87.3%
Coverage 27.5% 50.4%
Type of links HTML links RDF links
*Links manually generated by the WikiHow community
34. Know-How as Linked Data?
● It is possible to represent know-how as Linked Data
● Linked Data can be useful in the human know-how domain
for a large number of applications
● Experimental evidence has been provided in the use-case
of links generation
35. References:
Pareti Paolo, Ewan Klein, and Adam Barker. "A semantic web of know-how:
linked data for community-centric tasks." Proceedings of the companion
publication of the 23rd international conference on World wide web companion.
International World Wide Web Conferences Steering Committee, 2014.
Pareti Paolo, Ryutaro Ichise, Benoit Testu, Ewan Klein, and Adam Barker
“Integrating Know-How into the Linked Data Cloud” Knowledge Engineering and
Knowledge Management (EKAW 2014), 24 - 28 November 2014, Linköping,
Sweden (forthcoming)