WikiIndaba 2023/Submissions/Can we build machine translation technologies and LMs for written and oral African languages on Wikipedia that are respectful of communities’ knowledge, agency, and histories?

ID :
Author(s):	Username(s): Whose Knowledge?, Tinaral	Type of submission: Discussion
Author(s):	Affiliation: Whose Knowledge? User Group	Theme(s): Community Engagement, Language Justice, Technology
Abstract: Despite the hype behind so-called “artificial intelligence”, and the claims by big tech companies of how positive/essential/inevitable this machine learning technology is for our future, the truth is that digital language technologies in general, but especially the technologies relying on LLMs, are not constructed with the deep and dynamic social contexts and situated knowledges of language communities themselves, which means that they rarely embody the specific knowledge systems and practices of these communities. In this session we will explore what a community-first approach to the design and development of digital language technologies, especially machine translation and language models, would look like for written and oral African languages on Wikipedia.		Level of advancement: Basic
		Special requirements: Simultaneous interpretation into a local African language
		Extra information:

How will this session be beneficial for the communities in the region? The Wikipedia community is ramping up the implementation of digital language technologies relying on LLMs to make content in African languages more accessible. Oftentimes, these technologies are implemented without a deep power analysis on who is building it, who will benefit from it, who will have to face its ramifications, and who will decide its future. Exploring possibilities for designing and developing digital language technologies that are more fair, distributed, and community-based is crucial for the Wikimedia community at this point in time.

Interested participants edit

(register below and ask your questions now to the session organizer)