Localising Technical Hydroponics Concepts for Otjisa: A Software-Enabled Descriptive Translation Approach

The UNESCO Chair in Digital Technology Design with the Indigenous People cluster has worked with the Ovahimba community in Otjisa for many years to support community development through the co-design of technology and the introduction of emerging innovations. One such initiative involved a hydroponics system aimed at enhancing food and crop production. However, many technical hydroponics concepts do not have direct Otjiherero equivalents, making it challenging to teach community members how hydroponics works. Explaining concepts such as the nutrient film technique or pH levels becomes particularly difficult when the necessary terminology does not exist in the local language.

This qualitative research co-designed, developed, and evaluated a prototype community-based translation-support tool. The tool uses a two-stage Natural Language Processing pipeline consisting of a semantic simplification module for rephrasing English hydroponics concepts into clear, layman's English, and a translation module that renders these outputs into Otjiherero descriptions using descriptive translation strategies. The study employed Participatory Action Research methods, including participant observation, focus group discussions, role-play exercises, and technology probes with community members and researchers to co-design the mobile prototype.

Evaluation was conducted through thematic analysis and intrinsic metrics such as BLEU (0.5757) and METEOR (0.6237). The analysis revealed six interrelated themes, including Negotiating Meaning Across Languages and Culturally Embedded Teaching Practices. The resulting community translation-support tool displays a hydroponics-related image alongside a description generated by the system, when the image is tapped, a text-to-speech feature reads the Otjiherero description aloud to accommodate users with limited literacy. Findings confirmed the tool's utility for local learning while identifying a need for improved Text-to-Speech (TTS) pronunciation and grammatical accuracy in future iterations.