You are on page 1of 37

QUIPU:

Quechua Knowledge Graph


Pilot: Building a virtual assistant based on Quechua language
Elwin Huaman
August, 2019
What does Quipu mean?
or talking knots, Quipu [key-poo] is an ancient Incan
knowledge base and “writing system”, consisting of
various type and colours of knots tied to a main
cord, which represents both statistical (crops
grown, taxes, workers, mines, etc.) and narrative
information (stories and histories).

@ringmar.net
Seite 3
Seite 4
@interactive.quipu-project.com
“The global economy has been
transformed from a material-based
economy into a knowledge-based
economy. Whereas you can conquer oil
fields through war, you cannot acquire
knowledge that way. Hence today the
main source of wealth is knowledge.”
(Yuval Noah Harari)
Content

● Why QUIPU [The purpose]

● How we can built QUIPU [The process]

● What we achieve [The result]


Content

● Why QUIPU [The purpose]

● How we can built QUIPU [The process]

● What we achieve [The result]


Quechua speakers

@gemreportunesco.wordpress.com
@nephicode.blogspot.com
Ayllu (family): Huaman Quispe @quechuaexperience.github.io
Quechua education

@movehub.com @amazon.com
“The Web does
not just connect
machines, it
connects people”
(Tim Berners-Lee)

@123rf.com/profile_franzidraws
How does
language
shape your
experience of
the internet?

@internetsociety.org
Motivation
OK Google,
maypitaq Sorry, I
Machu Picchu cannot
kachkan? understand
you

Virtual assistant

Google Knowledge Graph


Solution
QUIPU,
maypitaq
Machu Picchu
Machu Picchu
kachkan Cusco
kachkan?
Suyupi

Virtual assistant

QUIPU: Quechua Knowledge Graph


Content

● Why QUIPU [The purpose]

● How we can built QUIPU [The process]

● What we achieve [The result]


Do we need the help of Machines?

Is information understandable by Humans and Machines?

Example: “Machu Picchu was built in c. 1450”

Machu Picchu can be:

Inca Citadel Town Store


https://www.wikidata.org/entity/Q676203 https://www.wikidata.org/entity/Q397990 https://www.wikidata.org/entity/Q2886434
Do we need the help of Machines?

How machines represent information?

Example: “Machu Picchu was built in c. 1450”

Simple statement: (Subject, Predicate, Object)

https://www.wikidata.org/prop/direct/P571
https://www.wikidata.org/entity/Q676203 "c. 1450"^^http://www.w3.org/1999/02/22-rdf-syntax-ns#langString
built
Machu Picchu c. 1450
Do we need the help of Machines?

How machines represent information?

Example: “Machu Picchu was built in c. 1450”

Simple statement: (Subject, Predicate, Object)


Prefix declarations:
wd :<https://www.wikidata.org/entity/>
wdp :<https://www.wikidata.org/prop/direct/>
rdf :<http://www.w3.org/1999/02/22-rdf-syntax-ns#>

wdp:P571
wd:Q676203 "c. 1450"^^rdf:langString
built
Machu Picchu c. 1450
Prefix declarations:
What is a wd
wdp
:<https://www.wikidata.org/entity/>
:<https://www.wikidata.org/prop/direct/>
Knowledge Graph? rdf
xsd
:<http://www.w3.org/1999/02/22-rdf-syntax-ns#>
:<http://www.w3.org/2001/XMLSchema#>

"c. 1450"^^rdf:langString
im 71
ag ilt :P5
e Machu Picchu Bu wdp
wd
p:P
...jpg 18
wd:Q676203 lo
ca
w tion Cusco
dp
:P
13 wd:Q5582862

6
1

259
wd re
p:P
po

tu
pu

cul
wd latio
p:P n
108
Built Inca Empire 2
wdp:P571 capital
wd:Q28573 wdp:P36
"c. 1438"^^rdf:langString 428 450^^xsd:integer
Prefix declarations:
What is Quechua wd
wdp
:<https://www.wikidata.org/entity/>
:<https://www.wikidata.org/prop/direct/>
Knowledge Graph? rdf
xsd
:<http://www.w3.org/1999/02/22-rdf-syntax-ns#>
:<http://www.w3.org/2001/XMLSchema#>

a "c. 1450"^^rdf:langString
h isk
i c 1
wa
nk t ar :P57
i Machu Pikchu Ha wdp
wd
p:P
...jpg 18
wd:Q676203 su
yu
w Qusqu
dp
:P
13 wd:Q5582862

6
1

259
wd ay
p:P
kaw

s
kaw
s
wd aqku
p:P n
108 a
Hatarichiska Tawantinsuyu 2
aqta
wdp:P571 umalli_ll
wd:Q28573 wdp:P36
"c. 1438"^^rdf:langString 428 450^^xsd:integer
How to represent a Knowledge Graph?

RDF (Resource Description Framework) allows to represent knowledge


graphs using syntaxes like Turtle, N-Triples, JSON-LD,...

prefix dbr :<http://dbpedia.org/resource/>


e.g. RDF model using Turtle prefix dbo :<http://dbpedia.org/ontology/>
prefix xsd :<http://www.w3.org/2001/XMLSchema#>

dbr:Peru dbo:longName “Republic of Peru”^^xsd:string ;


dbo:capital dbr:Lima ;
dbo:currency dbr:Peruvian_sol ;
dbo:demonym “Peruvian”^^xsd:string ;
dbo:populationTotal “31 151 643”^^xsd:integer .
dbr:Lima dbo:populationTotal “8852000”^^xsd:integer ;
dbo:country dbr:Peru .
How to build a
Knowledge graph

Talking Knowledge graphs: https://www.slideshare.net/STI-Innsbruck/talking-knowledge-graphs-ny


How to build a
Knowledge graph
Requirements:
● a well-known “standard” Ontology
or vocabulary, e.g. DBpedia Ontology

● homogeneous structure/models
e.g. a Place might be represented using
similar properties

● correct and complete information


e.g. how accurate is the knowledge

@ontology2.com
Knowledge Creation

Methods
● Manual, uses Annotation tool for an specific domain.

● Semi-automatic, uses intermediate tools for extraction( e.g. Crawlers)


and for mapping(e.g. Annotation Editor) information.

● Mapping, maps different formats to an specific ontology and integrates


large knowledge bases.

● Automatic, applies Natural Language Processing (NLP), Machine


Learning (ML), and more.
Knowledge Creation

Sources for creating QUIPU:


● RDF Exports from Wikidata
● Wikidata Toolkit
● Wikidata Sparql endpoint (export/consume)
● Wikidata ApiSandbox (search entities)
● Quechua Wiktionary
● Quechua Wikipedia
● Wikipedia Extractor
● Quechua Dictionary
● Microsoft Translator
Knowledge Creation

e.g. search antañiqiq entity.


Knowledge Hosting

Requirements:
● Annotation - Tool
e.g. A platform for creating and hosting annotations
● Document store for hosting semantic web annotations
e.g. MongoDB for hosting semantically annotated data based on JSON-LD.
● Graph Database for Hosting the Knowledge Graph
e.g. GraphDB for hosting semantically annotated data based on RDF.
Knowledge Curation
Entity

Literal

Relationship
Requirements:
● Assessment Machu Pikchu

i.e. assess the quality


● Cleaning um

a
a

sk
lo lli_

hi
i.e. assess the correctness ca lla

kawsay
ilt ri
tio qt

culture
bu ata
n a
● Enrichment

H
i.e. assess the completeness

c.1536
1450
Qusqu
a
i_llaqt
umall
l
Tawantinsuyu capita
Knowledge Deployment

Requirements: @slack.com @facebook.com @telegram.org

● Knowledge management technology


○ e.g. GraphDB
● Data accessibility
○ e.g. personalized agents
● Conversational interfaces
○ e.g. automating customer communication, chatbots
@amazon.com @google.com

Conversational user interfaces


(e.g. chatbots, voice assistants)

Talking Knowledge graphs: https://www.slideshare.net/STI-Innsbruck/talking-knowledge-graphs-ny


Content

● Why QUIPU [The purpose]

● How we can built QUIPU [The process]

● What we achieve [The result]


Pilot: QUIPU (Quechua Knowledge Graph)

● Knowledge Creation
○ e.g. Create a RDF model using Turtle syntax

prefix dbr :<http://dbpedia.org/resource/>


prefix dbo :<http://dbpedia.org/ontology/>
prefix xsd :<http://www.w3.org/2001/XMLSchema#>

dbr:Peru dbo:longName “Republic of Peru”^^xsd:string ;


dbo:capital dbr:Lima ;
dbo:currency dbr:Peruvian_sol ;
dbo:demonym “Peruvian”^^xsd:string ;
dbo:populationTotal “31 151 643”^^xsd:integer .
dbr:Lima dbo:populationTotal “8852000”^^xsd:integer ;
dbo:country dbr:Peru .
Pilot: QUIPU (Quechua Knowledge Graph)

● Knowledge Hosting
e.g. use GraphDB to store the knowledge graph

● Knowledge Curation
Assessment of the quality, using metrics
Cleaning, detect and correct errors
Enrichment, detect duplicates and resolve conflicting property values.
Pilot: QUIPU (Quechua Knowledge Graph)

● Knowledge Deployment
○ e.g. personalized agents
○ e.g. DialogFlow
○ e.g. MycroftAI
■ skill-Wiki
■ *develop a Quechua speech recognition skill
● based on a Spanish voice assistant
■ *use the Quechua Wikipedia
■ *develop skill-Quechua-Wiki
Pilot: QUIPU (Quechua Knowledge Graph)
Pilot: QUIPU (Quechua Knowledge Graph)
Take away
● facilitate sustainable development of cultural heritage knowledge in developing
countries through promoting technological support in a native language.
● increase access to information and communication technology in their native
language and will decrease the digital illiteracy.
● Reduce inequalities by given access to new technologies to indigenous communities,
it can ensure that the new decisions for developing technologies (e.g. interfaces) also
have to consider the Quechua language.
● Quality education (e.g. ensure that children and old people can acquire the knowledge
and skills needed in their native language)
Tupananchiskama! @qu
Thank you & see us! @en

[user:elwinlhq]

You might also like