Building Knowledge Graphs for Under Resourced Languages
Building Knowledge Graphs for Under Resourced Languages
Profile picture for user elwin.huaman
Elwin Huaman

Building Knowledge Graphs for Under Resourced Languages

Förderjahr 2024 / Stipendium Call #19 / Stipendien ID: 7335

This work aims to lay the foundations for building community-centred and language-aligned Knowledge Graphs for Under-Resourced Languages and their communities. These Knowledge Graphs will act as a common sense layer and will play a key role in shaping the Internet and Artificial Intelligence applications for a better future, where no one is left behind. Why this project matters: There are about 7000 languages in the world and only a few of them have been implemented in language technologies. How it is done: Allowing all languages to be represented in knowledge graphs Contributions: A community-driven approach to knowledge graph curation, and bringing awareness of bias in knowledge graphs and AI applications.

Uni | FH [Universität]

Universität Innsbruck

Themengebiet

Artificial Intelligence
,
Digital Divide
,
Information Visualization
,
Interdisciplinary Research
,
Projektmanagement
,
Semantic Web
,
Semantische Modellierung

Zielgruppe

Start-ups
,
Systemintegratoren
,
thematische Community

Gesamtklassifikation

Dissertation | PhD
,
Informationen | Guidelines
,
Initiative |Aktion

Technologie

AI | KI
,
Datenbank
,
Linked Data
,
Open Data
,
Semantische Netze

Projektergebnisse

Zwischenbericht CC-BY

Building Knowledge Graphs for Under-Resourced Languages Interim Report

stip7335_Call19_Zwischenbericht_V01.pdf
592.56 KB
Interim Report
Paper CC-BY-SA

Huaman, E., Huaman, J. L., Huaman, W., & Quispe, N. (2025). Quechua speech datasets in common voice: The case of Puno Quechua. In *Information management and big data – 12th annual international conference (SIMBig 2025), Lima, Peru, October 29-31, 2025, Proceedings. Springer.

https://arxiv.org/pdf/2510.13871
Quechua Speech Datasets in Common Voice: The Case of Puno Quechua
Datensammlung CCO

This datasheet is for version 2.0 of the the Mozilla Common Voice Spontaneous Speech dataset for Puno Quechua (qxp). The dataset contains 2211 clips representing 10 hours of recorded speech (6 hours validated) from 26 speakers.

https://datacollective.mozillafoundation.org/datasets/cmj8u48et004tnxzps28psruc
Common Voice Spontaneous Speech 2.0 - Puno Quechua
Datensammlung CCO

This datasheet is for version 24.0 of the the Mozilla Common Voice Scripted Speech dataset for Puno Quechua (qxp). The dataset contains 9033 clips representing 11.63 hours of recorded speech (9.88 hours validated) from 14 speakers.

https://datacollective.mozillafoundation.org/datasets/cmj8u3pq700ndnxxb737sq5re
Common Voice Scripted Speech 24.0 - Puno Quechua
Präsentation CC-BY-SA

I act as part of the advisory committee for the Shared Task: Mozilla Common Voice Spontaneous Speech ASR, that challenges researchers and developers to push ASR further, across 21 underrepresented languages from Africa, Asia, Europe and the Americas.

https://community.mozilladatacollective.com/shared-task-mozilla-common-voice-spontaneous-speech-asr/
Shared Task: Mozilla Common Voice Spontaneous Speech ASR

Blogbeiträge

Puno Quechua is an agglutinative language with nearly half a million speakers, yet it has faced a steady decline in usage over the decades. We explain how to use modern tools to revitalize its digital presence.

( )
Case Study: Puno Quechua language

Event in Puno about the creacion of speech dataset for Quechua communities.
I highlight the broad effort to bring 17 Quechua languages into Mozilla’s Common Voice platform to combat data scarcity. I outline how a community-driven, collaborative framework has successfully amassed 191.1 hours of speech data, with an 86% valida

( )
Giving a Voice to Quechua: Breaking Digital Barriers

Speech data collection event in Puno, Peru
Speech datasets empowers language communities to contribute significantly to the technology that will shape their future, turning speech AI into a powerful tool for preservation, revitalization, and modern connection.

( )
Unlocking the Future of Speech data for Under-Resourced Languages

Under Resource Languages Workshop
The digital space is dominated by a few major languages, creating barriers for speakers of under-resourced languages (URLs), who lack access to language technologies and information. Knowledge Graphs (KGs) have supported applications in well-resource

( )
Multilingual Awareness in Knowledge Graphs

What are Knowledge Graphs.
The digital space is dominated by a few languages, disadvantaging speakers of under-resourced languages (URLs). Knowledge graphs (KGs), which are used by major tech companies, have worked well for well known languages, however, the potential of KGs t

( )
Bridging the Digital Gap

Ähnliche Stipendien

Efficient data exchange between classical and quantum systems.

QUDAPI: Efficient Data Pipeline in Quantum-enhanced Cloud Computing

Advancing Privacy in Federated Learning

group of IoT devices
Research into the security of IoT devices and analysis of their vulnerabilities in order to better understand the “Internet of Insecure Things” and make it more secure.

Analyzing and Understanding the Internet of Insecure Things

Stiftbild

Intellectual Property Protection in Open Data Sharing

Dynamic analysis of mobile apps
Large-scale mobile app analysis to find security and privacy issues

Hidden Dangers: Uncovering Security and Privacy Risks through Large-scale Mobile App Analysis