/ PROJECT

Tokotoko — Te Reo Language Learning Application

/ DATE

2024—2025

/ ROLE

Design, development, research

/ TYPE

Thesis, participatory mapping, data viz

Exhibition installation image from Mapping Human Earth Systems

Tokotoko is a co-designed platform for everyday Te Reo Māori. It focuses on learning in lived contexts: short videos, conversations, and tasks that mirror real use, not drills. The project originated with project lead Jesse Pirini and a gap we observed between beginner tools and the needs of intermediate learners. We used cultural probes and field research to map when, where, and why learners engage, then translated those insights into product mechanics.

Design is guided by kaupapa Māori values and evidence-based pedagogy. Content is sequenced via a familiarity measure inspired by i+1 and Zipfian frequency, keeping material comprehensible while adding stretch. Interactive transcripts, instant vocab capture, and spaced review connect viewing to retention. Gamification privileges whanaungatanga and confidence building over leaderboards.

  • Everyday input: short, authentic clips with subtitles and word-level timing
  • Adaptive flow: recommendations tuned to familiarity and coverage
  • Community playlists: kaiako and learners curate themed paths
  • Ethical data: user control and mātauranga Māori data stewardship

Discovery. Cultural probe kit, diaries, and photo prompts to capture daily language touchpoints. UX surveys and interviews to prioritise barriers and motivators.

Prototype. Interactive transcript player, vocabulary capture, and spaced repetition. Early playlists and difficulty tagging. Low-latency UI tests on mobile.

Evaluate. Usability tests on comprehension, retention, and affect. Compare playlist vs. algorithmic feed. Measure time-on-task and return rate.

Iterate. Refine familiarity model, add social features, and extend content taxonomy for contexts, registers, and domains.

A lightweight web stack optimised for media and annotation. Focus on transparent data models and exportability.

  • Web app: Next.js + TypeScript
  • Transcripts: WebVTT with word-level timing and bilingual alignment
  • Storage: Postgres (Supabase) with RLS for user data control
  • Review engine: spaced repetition with item difficulty and lapse handling
  • Analytics: privacy-preserving event pipeline for learning metrics
  • Exports: CSV/JSON and Anki-ready decks

Dashboards focus on comprehension and exposure rather than vanity scores.

  • Familiarity graph: tracks coverage of high-frequency vocab and multiword phrases
  • Context heatmap: domains used this week (home, mahi, kai, travel)
  • Retention bands: new, learning, mature, and due items
  • Journey map: playlist paths and branching choices

The aim is durable confidence in everyday reo. We test whether contextual input plus transparent review improves continued use, not just short-term streaks. The research outputs include a co-design playbook, open schemas for transcript annotation, and evaluation of culturally aligned gamification patterns.

  • Next.js, TypeScript
  • Postgres/Supabase
  • WebVTT, FFmpeg
  • Figma for UX
  • Python notebooks for analysis