After a PSLE English Oral practice on PSLEPrep, your child sees marks for pronunciation, fluency, expression, accuracy (reading), and content and language use (conversation). This page explains how those numbers are produced, how they relate to the 2025 SEAB format, and what to do if a dimension shows “Not scored”.
Two kinds of evidence: audio and meaning
Oral assessment always splits into delivery (how it sounds) and substance (what was said). PSLEPrep uses speech analysis on the actual recording for reading-aloud delivery where the service is available, and an AI language model to interpret transcripts for feedback, content depth, and language quality — including the full Stimulus-Based Conversation at the end of the session.
Reading Aloud (15 marks): Azure pronunciation assessment + PACT-aware expression
From 2025, every Reading Aloud task includes a PACT-style preamble (Purpose, Audience, Context, Tone). PSLEPrep mirrors that: your child sees a preamble before the passage, then reads aloud.
When our English pronunciation pipeline is available, we run Microsoft Azure Speech pronunciation assessment on the recording, aligned to the passage text. That gives word-level evidence — articulation, stress, omissions, substitutions — rather than guessing from text alone.
Scores are mapped to the same four buckets tutors use for the new 15-mark reading paper: pronunciation, fluency, expression (including whether delivery fits the preamble), and accuracy. A second model pass writes English-only feedback and can apply a small expression adjustment when the transcript clearly supports stronger or weaker PACT awareness.
If audio analysis is down
We do not invent pronunciation, fluency, or expression marks. Those dimensions show “Not scored” and we rely on transcript comparison for accuracy only — same honesty bar as our Chinese oral flow.
Stimulus-Based Conversation (25 marks): structured questions, then holistic scoring
The in-examiner dialogue follows the 2025 pattern: picture-based observation, personal experience, then opinion. After the last answer, the model returns structured scores aligned to a 25-mark conversation rubric — pronunciation, fluency, content, and language use — plus per-question notes, highlights, and a parent tip. For tips on the observation phase, see our guide on 5W1H photograph analysis for PSLE English Oral.
Content and language use together carry most of the weight, reflecting SEAB's emphasis on reasons, examples, and clear opinions in the new all-opinion format. Pronunciation and fluency in this block are interpreted conservatively from the conversation transcript (and any supplementary signals we pass in), so students are not over-credited when the model cannot hear the raw audio for that phase.
How this maps to PSLE (and where we stop)
PSLEPrep totals 40 marks for English Oral practice — 15 reading + 25 conversation — matching the post-2025 paper weighting. SEAB does not publish fine-grained mark schemes; our splits follow the syllabus structure and standard teacher frameworks (including PEEL / PEERS-style coaching in feedback).
We are not SEAB. Scores are for practice, progress tracking, and targeted revision — not a guarantee of exam marks.
| Paper section | Marks | PSLEPrep approach |
|---|---|---|
| Reading Aloud | 15 | Azure audio assessment + PACT-aware expression; English feedback |
| Stimulus-Based Conversation | 25 | AI review of full conversation — content, language, delivery dimensions |
Free diagnostic vs full practice
The free diagnostic samples one conversation question and uses the same subject pipelines (including pronunciation assessment when available). Full sessions add the complete three-question arc, history, and detailed reading breakdown. Chinese oral scoring uses a different engine (tone-level analysis) but the same design principle: audio for sounds, models for meaning.
Ready for structured 2025-format practice? Start free trial →
Frequently asked questions
How does PSLEPrep score English pronunciation?
For Reading Aloud, we use Microsoft Azure Speech pronunciation assessment on your child's recording, compared to the passage text. That captures word-level accuracy, common Singapore issues (e.g. final consonants, 'th'), and fluency signals. If the service is unavailable, those dimensions are not guessed — they show as Not scored.
Are PSLEPrep English scores official PSLE marks?
No. We align sections and totals to the 2025 SEAB format (15 + 25 = 40) and standard rubric language, but we are not examiners. Use scores to track improvement and prioritise revision.
Why does conversation scoring mention transcript limits?
The live examiner experience is audio-first, but the end-of-conversation scorecard is generated from the full transcript (with conservative caps on delivery dimensions when audio context is limited). Content and language use — the largest levers for AL bands — come from what your child actually said.
What does Not scored mean?
The pronunciation pipeline failed or was disabled for that attempt. Retry the reading; when Azure is reachable again, you'll get the full 15-mark breakdown.
How should we practise PACT at home?
Use the same passage with three different imaginary preambles (different audience and tone) and record all three — the reads should sound clearly different. Our PACT explainer is linked below.