Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category...
Transcript of Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category...
![Page 1: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/1.jpg)
Improving Effectiveness of Knowledge Graph GenerationRule Creation and ExecutionPieter HeyvaertPhD advisors: Ruben Verborgh and Anastasia Dimou
1
![Page 2: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/2.jpg)
More and more devices and applications are connected to the Internet
Devices
2
![Page 3: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/3.jpg)
More and more devices and applications are connected to the Internet
Devices
Applications
3
![Page 4: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/4.jpg)
More and more data is generated
Tweets/min.(x 1000)
year4
![Page 5: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/5.jpg)
New techniques deployed using this data on the WebData can be analyzed, combined and shared.
→ Powerful, new techniques can be designed and deployed on the Web.
Examples areArtificial intelligence used by personal assistants (Siri, Alexa)Improved search engines (Google, Bing)
5
![Page 6: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/6.jpg)
OverviewMoving from Web to Semantic Web
Challenges
Contributions
Conclusions
6
![Page 7: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/7.jpg)
OverviewMoving from Web to Semantic Web
Challenges
Contributions
Conclusions
7
![Page 8: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/8.jpg)
Original data representations on the Web are insufficient to exchange and reuse dataOriginally data on the Web is represented using different data formats.
Extra data can be added without saying what this data means.
→ Makes it hard to exchange and reuse data on the Web.
8
![Page 9: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/9.jpg)
Original data representations on the Web are insufficient to exchange and reuse dataOriginally data on the Web is represented using different data formats.
Extra data can be added without saying what this data means.
→ Makes it hard to exchange and reuse data on the Web.
9
![Page 10: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/10.jpg)
Example: able to share data between same contacts apps
10
![Page 11: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/11.jpg)
Example: able to share data between same contacts apps
11
![Page 12: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/12.jpg)
Example: able to share data between same contacts apps
12
![Page 13: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/13.jpg)
Example: not able to share data between different contacts apps
13
![Page 14: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/14.jpg)
Example: not able to share data between different contacts apps
14
![Page 15: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/15.jpg)
Semantic Web allows to easily exchange and reuse data
15
![Page 16: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/16.jpg)
OverviewMoving from Web to Semantic Web
Knowledge graphsKnowledge graph generationKnowledge graph inconsistenciesProcessors
Challenges
Contributions
16
![Page 17: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/17.jpg)
The main Semantic Web technology is knowledge graphsKnowledge graphs = graphs that provide semantic descriptions of entities and their relationships.
17
![Page 18: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/18.jpg)
The main Semantic Web technology is knowledge graphsKnowledge graphs = graphs that provide semantic descriptions of entities and their relationships.
18
![Page 19: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/19.jpg)
Example: knowledge graph
19
![Page 20: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/20.jpg)
Example: data about a person
Professor Oakhas the name
20
![Page 21: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/21.jpg)
Example: data about a person
21
This is already a graph.
Professor Oakname
![Page 22: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/22.jpg)
Example: data about a person
Professor Oakname
location
22
![Page 23: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/23.jpg)
Example: data about his location
Pallet Town
23
![Page 24: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/24.jpg)
Example: data about his location
24
Pallet Town Laboratory
![Page 25: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/25.jpg)
Example:combining data about person and location
Professor Oakname
location
Laboratory
Pallet Towncity
type
25
![Page 26: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/26.jpg)
Knowledge graph provides semantic descriptions
Professor Oakname
location
Laboratory
Pallet Towncity
type
name
“The full name of a person.”
city
“The city where a building is located.”
26
Semantic descriptions give the data meaning
![Page 27: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/27.jpg)
The main Semantic Web technology is knowledge graphsKnowledge graphs = graphs that provide semantic descriptions of entities and their relationships.
27
![Page 28: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/28.jpg)
OverviewMoving from Web to Semantic Web
Knowledge graphsKnowledge graph generationKnowledge graph inconsistenciesProcessors
Challenges
Contributions
28
![Page 29: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/29.jpg)
Knowledge graphs are often generated from other data sources
Database
File
Knowledge graph
29
![Page 30: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/30.jpg)
Example: knowledge graphs are often generated from other data sources
Database
File
Knowledge graph
30
![Page 31: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/31.jpg)
Knowledge graphs are often generated via rules
Database
File
Knowledge graph
Rules
31
![Page 32: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/32.jpg)
Rules add semantic descriptions to data
Professor Oakname
“The full name of a person.”
32
![Page 33: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/33.jpg)
Example: rules for a person
id name place
oak Professor Oak lab
ash Ash Ketchum h1
Every row in the table is a person.
“name” is the official full name of a person.
“place” refers to a person’s current location.
33
![Page 34: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/34.jpg)
Example: rules for a location
id category city
lab laboratory Pallet Town
h1 house Pallet Town
h2 house Pallet Town
Every row in the table is a location.
“category” is the type of a location.
“city” refers to a location’s city it is in.
34
![Page 35: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/35.jpg)
Example: executing rules results in knowledge graph
id name place
oak Professor Oak lab
ash Ash Ketchum h1
id category city
lab laboratory Pallet Town
h1 house Pallet Town
h2 house Pallet Town Knowledge graph
Every row in the table is a person.
“name” is the offcial full name of a person.
“place” refers to a person’s location in other table.
Every row in the table is a location.
“category” is the type of a location.
“city” refers to a location’s city it is in.
35
![Page 36: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/36.jpg)
Syntax and grammar of rules defined byknowledge graph generation languageSyntax and grammar: how to create rules correctly.
Similar to natural languages, such as English and Dutch.
Examples of knowledge graph generation languages:R2RML: W3C recommendation for databases.RML: Extension of R2RML for multiple data sources in different formats.
36
![Page 37: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/37.jpg)
OverviewMoving from Web to Semantic Web
Knowledge graphsKnowledge graph generationKnowledge graph inconsistenciesProcessors
Challenges
Contributions
37
![Page 38: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/38.jpg)
Knowledge graphs can containtwo types of inconsistenciesNot using the suitable concepts and relationships.
Concepts and relationships do not model the real world as desired.
38
![Page 39: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/39.jpg)
Example: not using suitable concepts and relationships
Professor Oakage
location
Laboratory
Pallet Townstreet
type
39
![Page 40: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/40.jpg)
Example: concepts and relationships do not model real world as desired
Professor Oakname
location
Laboratory
Pallet Towncity
type
“A name has a number as value.”
40
![Page 41: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/41.jpg)
Possible causes for inconsistenciesRules: wrong concepts and relationships are used.
For example, using the relationship “age” instead of “name”.
Definitions of concepts and relationships: incorrectly model real world.For example, a name only consists of numbers.
Inconsistencies are fixed by updating the rules, definitions or both of them.
41
![Page 42: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/42.jpg)
OverviewMoving from Web to Semantic Web
Knowledge graphsKnowledge graph generationKnowledge graph inconsistenciesProcessors
Challenges
Contributions
42
![Page 43: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/43.jpg)
Processor executes rules to generate knowledge graphProcessor = software tool that, given a set of rules and data sources, generates knowledge graphs.
There exists at least one processor for each language:RMLMapper and RMLStreamer for RMLMorph-RDB and R2RML Parser for R2RML
43
![Page 44: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/44.jpg)
Switch between different processors
44
data sources knowledge graph
processor A
processor B
processor C
rules
Rules remain unchanged.
![Page 45: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/45.jpg)
Putting it all together
45
knowledge graph
rulesrule creation
rule execution
rule refinement
![Page 46: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/46.jpg)
46
![Page 47: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/47.jpg)
This already existed.I didn’t do anything here.
47
![Page 48: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/48.jpg)
OverviewMoving from Web to Semantic Web
Challenges
Contributions
Conclusions
48
![Page 49: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/49.jpg)
Three research challenges
49
knowledge graph
rulesrule creation
rule execution
rule refinement
Challenge 1
Challenge 2
Challenge 3
![Page 50: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/50.jpg)
Creating rules requires lot of human effortComponents: data sources, rules, concepts, and relationships.
Data sources can be in different formats (databases, files).
Data might need to be transformed (capitalize names).
Lot of rules might be needed.Rules are meant for machines, not humans.
Different concepts and relationships are used (how are they related).
50
![Page 51: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/51.jpg)
Challenge 1: make it easier to create rules
51
![Page 52: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/52.jpg)
Rules and definitions of concepts and relationships can lead to inconsistencies.Rules: wrong concept and relationships are used.
For example, using the relationship “age” instead of “name”.
Definitions of concept and relationships: incorrectly model real world.For example, a name only consists of numbers.
52
![Page 53: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/53.jpg)
Challenge 2: make it easier to fix inconsistencies in knowledge graphs
53
![Page 54: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/54.jpg)
Different processors have different featuresSpeed
Programming language
Follow syntax and grammar of knowledge graph generation language or not
54
![Page 55: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/55.jpg)
Challenge 3: make it easier to select the best processor
55
![Page 56: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/56.jpg)
OverviewMoving from Web to Semantic Web
Challenges
Contributions
Conclusions
56
![Page 57: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/57.jpg)
OverviewMoving from Web to Semantic Web
Challenges
ContributionsMapVOWLRMLEditorResglassRML test cases
Conclusions
57
![Page 58: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/58.jpg)
MapVOWL contributes to Challenge 1
58
knowledge graph
rulesrule creation
rule execution
rule refinement
Challenge 1
Challenge 2
Challenge 3MapVOWL
![Page 59: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/59.jpg)
MapVOWL: visual notation for knowledge graph generation rules
59
![Page 60: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/60.jpg)
Problem: visualizations not thoroughly investigated yet.Different tools have different graph-based visualizations.
No specification of visualizations specified → Hard to compare with others tools/visualizations.Hard to implement visualizations in other tools.
60
![Page 61: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/61.jpg)
MapVOWL: visual notation for knowledge graph generation rules
61
Graph-based
Models knowledge graphs
Colors to denote data sources
Independent of knowledge graph generation language
![Page 62: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/62.jpg)
Evaluation: compare MapVOWL with RML9 participants
List of questions answered for both MapVOWL and RML:Number of relationshipsNumber of data sourcesList relationships related to person
62
![Page 63: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/63.jpg)
Example: RML rules:map_anomaly_0 rml:logicalSource :source_0.
:source_0 a rml:LogicalSource;
rml:source "input.json";
rml:iterator "$";
rml:referenceFormulation ql:JSONPath.
:map_anomaly_0 a rr:TriplesMap;
rdfs:label "anomaly".
:s_0 a rr:SubjectMap.
:map_anomaly_0 rr:subjectMap :s_0.
:s_0 rml:reference "id".
:pom_0 a rr:PredicateObjectMap.
:map_anomaly_0 rr:predicateObjectMap :pom_0.
:pm_0 a rr:PredicateMap.
:pom_0 rr:predicateMap :pm_0.
:pm_0 rr:constant rdf:type.
:pom_0 rr:objectMap :om_0.
:om_0 a rr:ObjectMap;
rr:template
"http://IBCNServices.github.io/Folio-Ontology/Folio.owl#
{anomaly.type}";
rr:termType rr:IRI.
:pom_1 a rr:PredicateObjectMap.:map_anomaly_0 rr:predicateObjectMap :pom_1.:pm_1 a rr:PredicateMap.:pom_1 rr:predicateMap :pm_1.:pm_1 rr:constant dc:description.:pom_1 rr:objectMap :om_1.:om_1 a rr:ObjectMap;
rml:reference "anomaly.description";rr:termType rr:Literal.
:pom_2 a rr:PredicateObjectMap.:map_anomaly_0 rr:predicateObjectMap :pom_2.:pm_2 a rr:PredicateMap.:pom_2 rr:predicateMap :pm_2.:pm_2 rr:constant folio:hasCauseDescription.:pom_2 rr:objectMap :om_2.:om_2 a rr:ObjectMap;
rml:reference "anomaly.explanation";rr:termType rr:Literal.
:pom_3 a rr:PredicateObjectMap.:map_anomaly_0 rr:predicateObjectMap :pom_3.:pm_3 a rr:PredicateMap.:pom_3 rr:predicateMap :pm_3.:pm_3 rr:constant sosa:usedProcedure.:pom_3 rr:objectMap :om_3. 63
![Page 64: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/64.jpg)
Result: MapVOWL preferred over RMLCorrectness when answering questions is same for MapVOWL and RML.
Participants prefer MapVOWL over RML.
64
![Page 65: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/65.jpg)
BONUS: participants would like to use MapVOWL to edit rules.
65
![Page 66: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/66.jpg)
OverviewMoving from Web to Semantic Web
Challenges
ContributionsMapVOWLRMLEditorResglassRML test cases
Conclusions
66
![Page 67: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/67.jpg)
RMLEditor contributes to Challenge 1
67
knowledge graph
rulesrule creation
rule execution
rule refinement
Challenge 1
Challenge 2
Challenge 3MapVOWLRMLEditor
![Page 68: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/68.jpg)
RMLEditor: graph-based rule editor
68
![Page 69: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/69.jpg)
Problem: rule editors not thoroughly investigated yetDifferent approaches
Form-basedGraph-based
Support different data sourcesOnly databasesOnly files
(No) support for data transformations (e.g., capitalize names)
69
![Page 70: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/70.jpg)
RMLEditor’s GUI has three panelsInput Panel
Modeling Panel
Results Panel
70
![Page 71: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/71.jpg)
Input Panel shows data sources
71
![Page 72: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/72.jpg)
Modeling Panel allows to create rules via MapVOWL
72
![Page 73: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/73.jpg)
Results Panel shows generated knowledge graph
73
![Page 74: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/74.jpg)
Evaluation: compare RMLEditor with RMLx10 participants
Evaluation:Two use cases: either with RMLEditor or RMLxList of questions for RMLEditor
Large graphsData transformations
74
![Page 75: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/75.jpg)
Result: RMLEditor is better than RMLxHigher completeness of rules when using RMLEditor.
Participants rated RMLEditor as good, while RMLx as poor.
75
![Page 76: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/76.jpg)
Challenge 1MapVOWL + RMLEditor = easier to create rules.
76
![Page 77: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/77.jpg)
OverviewMoving from Web to Semantic Web
Challenges
ContributionsMapVOWLRMLEditorResglassRML test cases
Conclusions
77
![Page 78: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/78.jpg)
Resglass contributes to Challenge 2
78
knowledge graph
rulesrule creation
rule execution
rule refinement
Challenge 1
Challenge 2
Challenge 3MapVOWLRMLEditor
Resglass
![Page 79: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/79.jpg)
Resglass: rule-driven method to resolve inconsistencies
79
![Page 80: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/80.jpg)
Resglass: rule-driven method to resolve inconsistencies
Due to the lack of a screenshot here’s a dog riding a skateboard.
80
![Page 81: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/81.jpg)
Problem: removal of inconsistenciesIntroduced by rules.
Introduced by definitions of used concepts and relationships.
81
![Page 82: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/82.jpg)
Resglass consists of 5 steps1. Rule inconsistency detection
2. Rules and definitions refinement
3. Knowledge graph generation
4. Knowledge graph inconsistency detection
5. Rules and definitions refinement
82
![Page 83: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/83.jpg)
Step 1: Rules inconsistency detection
Person rules
Locationrules
Definitions
83
name
location
city
person
![Page 84: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/84.jpg)
Step 1: Rules inconsistency detection
Person rules
Locationrules
name
location
city
person
Name is not a number
Location is a person
Inconsistency
84
Definitions
![Page 85: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/85.jpg)
Step 1: Rules inconsistency detection
Person rules
Locationrules
name
location
city
person
Inconsistency
Rule/definition contributing to at least 1 inconsistency
85
Name is not a number
Location is a person
Definitions
![Page 86: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/86.jpg)
Step 2: Rules and definitions refinement2.1 Rules clustering
2.2 Rules and definitions ranking
2.3 Rules and definitions refinement
86
![Page 87: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/87.jpg)
Step 2.1: Rules clustering based on entityWhy? Rules concerning single entity impact each other → their refinements too.
Person rules
Locationrules
Rule contributing to at least 1 inconsistency
87
![Page 88: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/88.jpg)
Step 2.2: Rules and definitions ranking via scoreWhy? So we know which rules and definitions to inspect first.
Person rules
Locationrules
Rule contributing to at least 1 inconsistency
name
location
person
Ontologydefinitions
Ranking of rules Ranking of definitions
88
1.
2.
![Page 89: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/89.jpg)
Step 2.3: Rules and definitions refinementto resolve inconsistencies
Person rules
Locationrules
name
location
city
person
Ontologydefinitions
Rule contributing to at least 1 inconsistency
*
*Refinement*89
![Page 90: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/90.jpg)
Step 3: Knowledge graph generation
Database
File
Knowledge graph
Rules
90
![Page 91: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/91.jpg)
Step 4: Knowledge graph inconsistency detectionWhy? Certain inconsistencies cannot be detected via the rules alone.The complete knowledge graph is required.
Knowledge graph
Name is not capitalized.
“professor oak” instead of “Professor Oak”
91
![Page 92: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/92.jpg)
Step 5: Rules and definitions refinement
Person rules
Locationrules
name
location
city
personRule contributing to at least 1 inconsistency
*
Refinement*
92
Definitions
![Page 93: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/93.jpg)
Result: knowledge graph without inconsistencies
Database
File
Knowledge graph
Rules
93
![Page 94: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/94.jpg)
Result: knowledge graph without inconsistencies
Database
File
Knowledge graph
Rules
94
![Page 95: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/95.jpg)
Evaluation: compare ranking of Resglass with manual ranking of experts3 participants
Evaluation: Participants rank rules and definitions for manually.Comparison between ranking of Resglass and participants’ rankings.
95
![Page 96: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/96.jpg)
Result: Resglass can help experts80% overlap with ranking of experts for rules and definitions.
96
![Page 97: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/97.jpg)
Challenge 2Resglass makes it easier to fix inconsistencies.
97
![Page 98: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/98.jpg)
OverviewMoving from Web to Semantic Web
Challenges
ContributionsMapVOWLRMLEditorResglassRML test cases
Conclusions
98
![Page 99: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/99.jpg)
RML test cases contributes to Challenge 3
99
knowledge graph
rulesrule creation
rule execution
rule refinement
Challenge 1
Challenge 2
Challenge 3MapVOWLRMLEditor
Resglass
RML test cases
![Page 100: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/100.jpg)
Example: test caseApplication that calculates the sum of two numbers: sum
sum(x, y)
x + y
Test casessum(1, 1) = 2sum(2, 2) = 4sum(1, 2) = 3
If all these cases succeed then the application works correctly.
100
![Page 101: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/101.jpg)
Test cases to determine conformance of processors to languages
101
Test if correct knowledge graph is generated from one database.
Test if correct knowledge graph is generated from one file.
Test if correct knowledge graph is generated from two files.
rules
rules
rules
![Page 102: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/102.jpg)
Problem: what are the characteristics of test cases for knowledge graphs languages that support different data sources?No test cases for such languages (RML).
Only test cases for databases (R2RML).
102
![Page 103: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/103.jpg)
Create initial test cases for RMLRML supports different data sources.
No test cases existed for RML.
Remember that RML is an extension of R2RML.
RML test cases based on R2RML test cases.
103
![Page 104: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/104.jpg)
Benefits of RML test cases for developers and usersDevelopers determine how conformant processors are to RML specification.
Users can use test case results to select most suitable processor for use case.
104
![Page 105: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/105.jpg)
Challenge 3The RML test cases make it easier to select the best RML processor.
105
![Page 106: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/106.jpg)
OverviewMoving from Web to Semantic Web
Challenges
Contributions
Conclusions
106
![Page 107: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/107.jpg)
OverviewMoving from Web to Semantic Web
Challenges
Contributions
ConclusionsImpact of contributionsRemaining challenges and future directions
107
![Page 108: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/108.jpg)
Impact of contributions
108
knowledge graph
rulesrule creation
rule execution
rule refinement
MapVOWL
RMLEditor
Resglass
RML test cases
![Page 109: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/109.jpg)
MapVOWL and RMLEditor → Challenge 1
109
knowledge graph
rulesrule creation
rule execution
rule refinement
MapVOWL
RMLEditor
Resglass
RML test casesChallenge 1
Challenge 1: make it easier to create rules.
![Page 110: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/110.jpg)
Resglass→ Challenge 2
110
knowledge graph
rulesrule creation
rule execution
rule refinement
MapVOWL
RMLEditor
Resglass
RML test casesChallenge 1
Challenge 2: make it easier to fix inconsistencies in knowledge graphs.
Challenge 2
![Page 111: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/111.jpg)
RML test cases → Challenge 3
111
knowledge graph
rulesrule creation
rule execution
rule refinement
MapVOWL
RMLEditor
Resglass
RML test casesChallenge 1
Challenge 3: make it easier to select the best processor.
Challenge 2
Challenge 3
![Page 112: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/112.jpg)
OverviewMoving from Web to Semantic Web
Challenges
Contributions
ConclusionsImpact of contributionsRemaining challenges and future directions
112
![Page 113: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/113.jpg)
Rule creation: YARRRMLSome users prefer text-based solution over visualizations.
Solution: YARRRML
113
![Page 114: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/114.jpg)
Rule refinement: Resglass can further assist users in resolving inconsistencies
Suggest possible resolutions based on used and other concepts and relationships.
Suggest possible resolutions based on rules of other use cases.
114
![Page 115: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/115.jpg)
Rule execution: define abstract test cases for languagesBeyond RML
What needs to be tested?
115
![Page 116: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/116.jpg)
Define what is part of knowledge graph languageAnd maybe more importantly what is not.
Pre-processingPost-processingUse case-specific functions
Beyond RML
But RML (and its test cases) can serve as inspiration.
116
![Page 117: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/117.jpg)
Improving Effectiveness of Knowledge Graph GenerationRule Creation and ExecutionPieter HeyvaertPhD advisors: Ruben Verborgh and Anastasia Dimou
117
![Page 118: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/118.jpg)
118
![Page 119: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/119.jpg)
OverviewMoving from Web to Semantic Web
Knowledge graphsOntologyKnowledge graph generationKnowledge graph inconsistenciesProcessors
Challenges
Research questions and hypotheses
Contributions119
![Page 120: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/120.jpg)
Large graphs become complexDifficult to keep overview of rules.
Difficult to find specific rules.
Difficult to edit rules.
120
![Page 121: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/121.jpg)
Solution RMLEditor: 5 detail levelsHighest
High
Moderate
Low
Lowest
121
![Page 122: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/122.jpg)
Solution RMLEditor: 5 detail levelsHighest
High
Moderate
Low
Lowest
122
![Page 123: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/123.jpg)
Solution RMLEditor: 5 detail levelsHighest
High
Moderate
Low
Lowest
123
![Page 124: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/124.jpg)
Solution RMLEditor: 5 detail levelsHighest
High
Moderate
Low
Lowest
124
![Page 125: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/125.jpg)
Solution RMLEditor: 5 detail levelsHighest
High
Moderate
Low
Lowest
125
![Page 126: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/126.jpg)
Solution RMLEditor: 5 detail levelsHighest
High
Moderate
Low
Lowest
126
![Page 127: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/127.jpg)
Data values might need to be transformedOriginal name: professor oak
Preferred name: Professor Oak
Data transformation needed to fix the capitalization.
127
![Page 128: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/128.jpg)
Solution RMLEditor: define transformation in Input Panel
128
![Page 129: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/129.jpg)
Solution RMLEditor: define transformation in Input Panel
129
![Page 130: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/130.jpg)
Solution RMLEditor: define transformation in Input Panel
Transformation to use
Values to use
130
![Page 131: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/131.jpg)
Use Resglass’ ranking for (semi-)automatic solutionRanking is use case-specific.
(Semi)-automatic solution can do better than using a predefined set of refinements.
131
![Page 132: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/132.jpg)
Evaluation: compare RMLMapper and CARMLRML processors:
RMLMapperCARML
Ran RML test cases against both processors.
Test case results are eitherPassedFailedInapplicable: not implemented → not tested
132
![Page 133: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/133.jpg)
RMLMapper passes more test cases than CARML
RMLMapper CARML
Passed 277 84
Failed 20 33
Inapplicable 0 180
133
![Page 134: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/134.jpg)
RMLMapper and CARML are not perfect
RMLMapper CARML
Passed 277 84
Failed 20 33
Inapplicable 0 180
RMLMapper fails for some database test cases: automatic typing
CARML fails for core language aspects
134
![Page 135: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/135.jpg)
CARML does not support databases
RMLMapper CARML
Passed 277 84
Failed 20 33
Inapplicable 0 180
135
CARML does not support databases.
The RMLMapper does.
![Page 136: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/136.jpg)
More research to create new test casesSpecific of hierarchical data formats (JSON, XML)
Impact of data source language (SQL, JSONPath, XPath)
136
![Page 137: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/137.jpg)
Remaining challenges and future directionsRule creation: are visualizations the only way?
Rule refinement: What can we automate? Can we avoid inconsistencies?
Rule execution: What is really part of the test cases? What is really part of the language?
137
![Page 138: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/138.jpg)
Example: data about a person
Professor Oakhas the name
subject predicate object
138
![Page 139: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/139.jpg)
Arrows denote the subject and object of a relationship.
Knowledge graph is directed
Professor Oakname
location
Laboratory
Pallet Towncity
type
139
![Page 140: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/140.jpg)
Knowledge graph is directed
Professor Oakname
location
Laboratory
Pallet Towncity
type
Arrows denote the subject and object of a relationship.
subject
object
140
![Page 141: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/141.jpg)
Ontology includes machine-understandable definitions of concepts and relationshipsConcepts and relationships model real world.
For example: location, name, city
Their definitions are written in machine-understandable way.
Definitions are part of an ontology.
→ Applications can work with these concepts and relationships.
141
![Page 142: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/142.jpg)
Improving Effectiveness of Knowledge Graph Generation Rule Creation and Execution
142
![Page 143: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/143.jpg)
Summary of MapVOWLResearch Question 1: How can we design visualizations that improve the cognitive effectiveness of visual representations of knowledge graph generations rules?
MapVOWL is visual notation for knowledge graph generation rules.
Hypothesis 1: MapVOWL improves the cognitive effectiveness of the generation rule visual representation to generate knowledge graphs compared to using RML directly.
MapVOWL is preferred over RML.
143
![Page 144: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/144.jpg)
Summary of ResglassResearch Question 3: How can we score and rank rules and ontology terms for inspection to improve the manual resolution of inconsistencies?
RMLEditor is a rule-driven method to resolve inconsistencies.
Hypothesis 3: The automatic inconsistency-driven ranking of Resglass improves, compared to a random ranking, by at least 20% the overlap with experts’ manual ranking.
Resglass can improve resolution of inconsistencies, partly due to its ranking.
144
![Page 145: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/145.jpg)
Differences between R2RML and RML on test case-level
R2RML RML
Data errors Process stops Process doesn’t stop: best effort
Inverse expressions: to optimize execution
Tested, but can’t see difference in output
Not tested
SQL-specific features Tested Depending on data source language
Null values Tested Not supported for CSV and XML
Spaces in columns Tested Not supported for XML
145
![Page 146: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/146.jpg)
Summary of RML test casesResearch Question 4: What are the characteristics of test cases for processors that generate knowledge graphs from heterogeneous data sources independent of languages’ specifications?
Databases vs other data sources
Tabular vs hierarchical data formats
SQL vs other data source languages
RML test cases help define and explore these characteristics.
146
![Page 147: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/147.jpg)
Overview of contributions
knowledge graph
rulesrule creation
rule execution
rule refinement
MapVOWL
RMLEditor
Resglass
RML test cases
147
![Page 148: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/148.jpg)
Highest levelShows everything
148
![Page 149: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/149.jpg)
High levelShows everything
ExceptDatatype of literalLanguage of literal
149
![Page 150: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/150.jpg)
Moderate levelShows everything
ExceptDatatype of literalLanguage of literal
Relationships on arrows
150
![Page 151: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/151.jpg)
Low levelShows everything
ExceptDatatype of literalLanguage of literal
Relationships on arrows
Relationships that are not between two entities
151
![Page 152: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/152.jpg)
Lowest levelShows everything
ExceptDatatype of literalLanguage of literal
Relationships on arrows
Relationships that are not between two entities
Entities without an links152
![Page 153: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/153.jpg)
Summary of RMLEditorResearch Question 2: How can we visualize the components of a knowledge graph generation process to improve its cognitive effectiveness?
RMLEditor is graph-based rule editor.
Hypothesis 2: The cognitive effectiveness provided by the RMLEditor’s GUI improves the user’s performance during the knowledge graph generation process compared to RMLx.
RMLEditor is preferred over RMLx.
153
![Page 154: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/154.jpg)
Overview of contributions
knowledge graph
rulesrule creation
rule execution
rule refinement
MapVOWL
RMLEditor
Resglass
RML test cases
154
![Page 155: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/155.jpg)
Example: rules for a person and location combinedid full-name place
001 Professor Oak lab
Every row in the table is a person.
“name” is the official full name of a person.
“place” refers to a person’s current location in other table.
id category city
lab Laboratory Pallet Town
h1 house Pallet Town
h2 house Pallet Town
Every row in the table is a location.
“category” is the type of a location.
“city” refers to a location’s city it is in.
155
![Page 156: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/156.jpg)
Example: RML rules:map_anomaly_0 rml:logicalSource :source_0.
:source_0 a rml:LogicalSource;
rml:source "input.json";
rml:iterator "$";
rml:referenceFormulation ql:JSONPath.
:map_anomaly_0 a rr:TriplesMap;
rdfs:label "anomaly".
:s_0 a rr:SubjectMap.
:map_anomaly_0 rr:subjectMap :s_0.
:s_0 rml:reference "id".
:pom_0 a rr:PredicateObjectMap.
:map_anomaly_0 rr:predicateObjectMap :pom_0.
:pm_0 a rr:PredicateMap.
:pom_0 rr:predicateMap :pm_0.
:pm_0 rr:constant rdf:type.
:pom_0 rr:objectMap :om_0.
:om_0 a rr:ObjectMap;
rr:template
"http://IBCNServices.github.io/Folio-Ontology/Folio.owl#
{anomaly.type}";
rr:termType rr:IRI.
:pom_1 a rr:PredicateObjectMap.:map_anomaly_0 rr:predicateObjectMap :pom_1.:pm_1 a rr:PredicateMap.:pom_1 rr:predicateMap :pm_1.:pm_1 rr:constant dc:description.:pom_1 rr:objectMap :om_1.:om_1 a rr:ObjectMap;
rml:reference "anomaly.description";rr:termType rr:Literal.
:pom_2 a rr:PredicateObjectMap.:map_anomaly_0 rr:predicateObjectMap :pom_2.:pm_2 a rr:PredicateMap.:pom_2 rr:predicateMap :pm_2.:pm_2 rr:constant folio:hasCauseDescription.:pom_2 rr:objectMap :om_2.:om_2 a rr:ObjectMap;
rml:reference "anomaly.explanation";rr:termType rr:Literal.
:pom_3 a rr:PredicateObjectMap.:map_anomaly_0 rr:predicateObjectMap :pom_3.:pm_3 a rr:PredicateMap.:pom_3 rr:predicateMap :pm_3.:pm_3 rr:constant sosa:usedProcedure.:pom_3 rr:objectMap :om_3. 156
![Page 157: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/157.jpg)
OverviewMoving from Web to Semantic Web
Challenges
Research questions and hypotheses
Contributions
Conclusions
157
![Page 158: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/158.jpg)
158
knowledge graph
rulesrule creation
rule execution
rule refinement
Challenge 1
Challenge 2
Challenge 3
![Page 159: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/159.jpg)
Different tools to create rulesDifferent approaches
FormsGraph-based visualizations
Support different data formatsOnly databasesOnly files
(No) support for data transformations (e.g., capitalize names)
159
![Page 160: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/160.jpg)
Design of tools and GUIs not thoroughly investigated
160
![Page 161: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/161.jpg)
Research Question 1:How can we design visualizations that improve the cognitive effectiveness of visual representations of knowledge graph generation rules?
Human words: how can we make the visualizations better?
161
![Page 162: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/162.jpg)
Introduce MapVOWL to address Research Question 1Visual notation for knowledge generation rules
162
![Page 163: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/163.jpg)
Hypothesis 1:MapVOWL improves the cognitive effectiveness of the generation rule visual representation to generate knowledge graphs compared to using RML directly.
163
![Page 164: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/164.jpg)
Research Question 2:How can we visualize the components of a knowledge graph generation process to improve its cognitive effectiveness?
Human words: how can we make the GUIs better?
164
![Page 165: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/165.jpg)
Hypothesis 1:MapVOWL improves the cognitive effectiveness of the generation rule visual representation to generate knowledge graphs compared to using RML directly.
165
Human words: we hope MapVOWL is better than RML.
![Page 166: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/166.jpg)
Introduce the RMLEditorto address Research Question 2Graph-based rule editor
166
![Page 167: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/167.jpg)
Hypothesis 2:The cognitive effectiveness provided by the RMLEditor’s GUI improves the user’s performance during the knowledge graph generation process compared to RMLx.
167
![Page 168: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/168.jpg)
Hypothesis 2:The cognitive effectiveness provided by the RMLEditor’s GUI improves the user’s performance during the knowledge graph generation process compared to RMLx.
168
Human words: we hope the RMLEditor is better than RMLx.
![Page 169: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/169.jpg)
169
knowledge graph
rulesrule creation
rule execution
rule refinement
Challenge 1
Challenge 2
Challenge 3
![Page 170: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/170.jpg)
Resolving inconsistencies is not straightforwardUse case: DBpedia (knowledge graph generated from Wikipedia)
Total # rules: > 1,200,000# inconsistencies: > 2,000# rules involved in at least 1 inconsistency: > 1,300
Where do we start? Which rules do we check first? Which inconsistencies do we check first?
170
![Page 171: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/171.jpg)
Research Question 3:How can we score and rank rules and ontology terms for inspection to improve the manual resolution of inconsistencies?
Human words: where do we start when fixing the problems?
171
![Page 172: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/172.jpg)
Introduce Resglassto address Research Question 3Rule-driven method for the resolution of inconsistencies in rules and ontologies.
172
![Page 173: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/173.jpg)
Introduce Resglassto address Research Question 3Rule-driven method for the resolution of inconsistencies in rules and ontologies.
173
Due to the lack of a screenshot here’s a dog riding a skateboard.
![Page 174: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/174.jpg)
Hypothesis 3:The automatic inconsistency-driven ranking of Resglass improves, compared to a random ranking, by at least 20% the overlap with experts’ manual ranking.
174
![Page 175: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/175.jpg)
Hypothesis 3:The automatic inconsistency-driven ranking of Resglass improves, compared to a random ranking, by at least 20% the overlap with experts’ manual ranking.
175
Human words: we hope Resglass is better than doing something random.
![Page 176: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/176.jpg)
176
knowledge graph
rulesrule creation
rule execution
rule refinement
Challenge 1
Challenge 2
Challenge 3
![Page 177: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/177.jpg)
Test cases to determine conformance of processors to languagesTest case = a set of actions executed to verify a particular feature or functionality of the software application.
177
![Page 178: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/178.jpg)
Test cases to determine conformance of processors to languagesTest case = a set of actions executed to verify a particular feature or functionality of the software application.
178
![Page 179: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/179.jpg)
No test cases for all languagesOnly for languages that work with databases.
No for languages that work with different data sources and formats.
→ Processors that work with different data sources and formats not tested.
179
![Page 180: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/180.jpg)
Research Question 4:What are the characteristics of test cases for processors that generate knowledge graphs from heterogeneous data sources independent of languages’ specifications?
Human words: what is part of a test case and what not?
180
![Page 181: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/181.jpg)
Introduce initial set of conformances test cases for RML to address Research Question 4Databases (like R2RML)
CSV files
XML files
JSON files
181
![Page 182: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/182.jpg)
Overview of contributions
knowledge graph
rulesrule creation
rule execution
rule refinement
MapVOWL
RMLEditor
Resglass
RML test cases
182
![Page 183: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/183.jpg)
Adding meaning in practiceKnowledge graph consists of zero or more triples.
One triple consists of a subject, predicate, and object.
Subject and predicate and object have explicit meaning defined.
183
![Page 184: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/184.jpg)
Example: triple
Professor Oakhas the name
subject predicate object
184
![Page 185: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/185.jpg)
Subjects are in most cases linkshttps://google.com/
https://ugent.be
https://kanto.edu/oak
https://kanto.map/lab
185
![Page 186: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/186.jpg)
Why links?Uniquely identify a resource.
Possible to follow a link → get more information.
186
![Page 187: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/187.jpg)
Predicates are linksname → http://schema.org/name
location → http://schema.org/location
city → http://dbpedia.org/ontology/city
187
![Page 188: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/188.jpg)
Objects are in most cases links and literalsLinks
https://kanto.edu/oakhttps://kanto.map/lab
LiteralsText: “Professor Oak”, “Pallet Town”Numbers: 1, 100, 1000
188
![Page 189: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/189.jpg)
Example: triple
Professor Oakhas the name
https://kanto.edu/oak http://schema.org/name “Professor Oak”
189
![Page 190: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/190.jpg)
Tools to create rules: Map-On
190
![Page 191: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/191.jpg)
Tools to create rules: Juma
191
![Page 192: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/192.jpg)
Tools to create rules: RMLx
192
![Page 193: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/193.jpg)
Example: rules for person via MapVOWL
193
![Page 194: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/194.jpg)
Example: rules for person via MapVOWLid full-name place
oak Professor Oak lab
Every row in the table is a person.
Link of person is https://kanto.edu/ + id.
“full-name” is the name of a person.
“place” refers to a person’s location.
http://schema.org/Person
194
![Page 195: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/195.jpg)
Example: rules for person via MapVOWLid full-name place
oak Professor Oak lab
Every row in the table is a person.
Link of person is https://kanto.edu/ + id.
“full-name” is the name of a person.
“place” refers to a person’s location.
https://kanto.edu/{id}http://schema.org/Person
195
![Page 196: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/196.jpg)
Example: rules for person via MapVOWLid full-name place
oak Professor Oak lab
Every row in the table is a person.
Link of person is https://kanto.edu/ + id.
“full-name” is the name of a person.
“place” refers to a person’s location.
full-namehttps://kanto.edu/{id}http://schema.org/Person
196
![Page 197: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/197.jpg)
Example: rules for person via MapVOWLid full-name place
oak Professor Oak lab
Every row in the table is a person.
Link of person is https://kanto.edu/ + id.
“full-name” is the name of a person.
“place” refers to a person’s location.
full-namehttps://kanto.edu/{id}http://schema.org/Person
http://schema.org/name
197
![Page 198: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/198.jpg)
Example: rules for person via MapVOWLid full-name place
oak Professor Oak lab
Every row in the table is a person.
Link of person is https://kanto.edu/ + id.
“full-name” is the name of a person.
“place” refers to a person’s location.
full-namehttps://kanto.edu/{id} http://schema.org/name
placehttp://schema.org/location
http://schema.org/Person
198
![Page 199: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/199.jpg)
Example: rules for location via MapVOWLid category city
lab Laboratory Pallet Town
Every row in the table is a location.
Link of location is https://kanto.map/ + id.
“category” is the type of a location.
“city” refers to a location’s city it is in.
http://schema.org/Place
199
![Page 200: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/200.jpg)
Example: rules for location via MapVOWLid category city
lab Laboratory Pallet Town
Every row in the table is a location.
Link of location is https://kanto.map/ + id.
“category” is the type of a location.
“city” refers to a location’s city it is in.
https://kanto.map/{id}http://schema.org/Place
200
![Page 201: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/201.jpg)
Example: rules for location via MapVOWLid category city
lab Laboratory Pallet Town
Every row in the table is a location.
Link of location is https://kanto.map/ + id.
“category” is the type of a location.
“city” refers to a location’s city it is in.
categoryhttps://kanto.map/{id} rdf:type
http://schema.org/Place
201
![Page 202: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/202.jpg)
Example: rules for location via MapVOWLid category city
lab Laboratory Pallet Town
Every row in the table is a location.
Link of location is https://kanto.map/ + id.
“category” is the type of a location.
“city” refers to a location’s city it is in.
categoryhttps://kanto.map/{id} rdf:type
cityhttp://dbpedia.org/ontology/city
http://schema.org/Place
202
![Page 203: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/203.jpg)
Example: no link between persons and locations
full-namehttps://kanto.edu/{id} http://schema.org/name
placehttp://schema.org/location
http://schema.org/Person
categoryhttps://kanto.map/{id} rdf:type
cityhttp://dbpedia.org/ontology/city
http://schema.org/Place
203
![Page 204: Improving Effectiveness of Knowledge Graph Generation PhD ... · ash Ash Ketchum h1 id category city lab laboratory Pallet Town h1 house Pallet Town h2 house Pallet Town Knowledge](https://reader034.fdocuments.us/reader034/viewer/2022051918/600a6ef46a0c9b7d63183740/html5/thumbnails/204.jpg)
Example: add link between persons and locations
full-namehttps://kanto.edu/{id} http://schema.org/name
http://schema.org/location
http://schema.org/Person
categoryhttps://kanto.map/{id} rdf:type
cityhttp://dbpedia.org/ontology/city
http://schema.org/Place
place
204