How Edspark diagnoses, develops, and tracks reading progress in primary schools

Ask six or seven teachers to grade the same student's reading, and you'll often get six or seven different answers. Not because anyone is doing it wrong — but because reading level is genuinely hard to pin down consistently, especially across a busy timetable and dozens of students. It's the exact problem our primary school platform was built to solve: give every student the same consistent, objective assessment, based on how they actually read and what they actually understand.

Starting from the first 100 words

Instead of guessing a starting point, the assessment begins with the first 100 words of a passage — no random sampling. From there, a four-question framework escalates in difficulty progressively, so the system isn't just checking whether a student can decode words, but whether they actually understand what they've read.

• Q1 — Factual: what happened in the story, based on the first 100 words
• Q2 — Referencing: content outside the first 100 words, requiring wider reading
• Q3 — Thinking: connecting earlier content, requiring whole-passage understanding
• Q4 — Inferencing: reading between the lines, requiring full comprehension

The target accuracy zone is 75–85% — enough to confirm a student's true level without the test being trivially easy or punishingly hard. From there, the system auto-advances or drops a level based on performance, with no teacher guesswork required. If a student can't read the first 100 words fluently, there's no need to ask the comprehension questions at all — it's a direct drop, so nobody's time is wasted on a test they were never going to pass.

Reading aloud, then explaining it back

Reading level is only half the picture. Our oral development component is inspired by the Feynman Technique: students read a passage aloud, then explain what they learned in full sentences. The AI listens, evaluates, and responds — it never just delivers a monologue back at the student.

• Real-time feedback on pronunciation accuracy and speaking pace
• Self-correct marking — a student who catches their own mistake isn't penalised for it
• A deliberate 0.5–1 second delay, so the system listens before it judges
• Questions read aloud to support younger learners who aren't yet confident readers themselves
• Support for both English and Chinese (Cantonese) oral practice
• Dialogue mode, not monologue — the AI asks follow-up questions instead of just scoring and moving on

Built so students want to come back tomorrow

A platform that makes struggling readers feel worse about reading isn't one anyone will use for long. So low-performing students never see their raw scores — the experience leads with encouragement, not judgement. Cartoon UI and animated rewards mark the moment a student levels up, a digital certificate is generated on completion, and parents get an email notification the moment their child advances, transcript included.

For younger learners, a lot of thought went into small things that add up: a swipe or large page-turn button instead of fiddly navigation, and volume calibration handled through interactive animations — like an ABC song or a cartoon jump — rather than an intimidating settings menu.

One platform, three subjects

Beyond English reading, the same underlying framework extends to content-based oral assessment across core primary subjects, using one unified rubric so teachers only have to learn a single system:

• English — reading level assessment, oral development, and comprehension across the full P1–P6 journey, mapped across 30 levels and 60 books
• Chinese (中文) — Cantonese oral evaluation via Read Aloud, Dialogue, and Group Discussion, with students expressing what they learned in full, structured sentences
• General Studies & Science — content-based oral Q&A aligned to passage topics, where students explain concepts back rather than just recognising them

What teachers and parents actually see

A teacher dashboard shows every student's reading level, comprehension accuracy, and oral ability in one place, making it easy to identify students with the biggest gaps for follow-up during the recheck week. Parents get a clear, accessible progress summary in Traditional Chinese — again, without raw scores for students who are struggling, just progress and encouragement. And because the system flags in real time which students haven't completed a session, teachers can step in without needing to manually track every student's status themselves.

Where things stand right now

We're currently piloting this system with two primary schools in Hong Kong, with a full launch targeted for September 2026. It's early — which is exactly why we're running a real pilot instead of a big rollout, checking that the reading levels, the oral scoring, and the encouragement-first experience all hold up with real students before anyone else sees them.