Writing · ELLIPSE
ρ .846
Score correlation
N=100. Evidence for writing on the Standard scale. Correlation is not a percentage of exact agreement.
The methodology behind the assessment
A score should
show its work.
How is my child’s work actually graded?
RALE connects student work to evidence, feedback and a rubric score. A teacher reviews the assessment before it becomes a published report.
Follow the assessmentFrom a real report / Precision AuditView full excerpt ↗
Read across the rowThe original.
The original.
The correction.
The reason.
The student’s words stay beside the feedback. The skill code identifies what to work on next.
GRAM / AGREEMENT / SVA- 01Student work
- 02Evidence
- 03Skill-tagged feedback
- 04Rubric score
- 05Teacher review
- 06Report
RALE · Rubric Assessment Learning Engine · Writing, speaking, reading and listening inside Skyonomy.
The assessment domains
Different work.
Different evidence.
Writing
From the student’s sentence to a correction you can inspect.
Read the methodSpeaking
Spoken evidence, a transcript, and a reviewable judgement.
Read the methodReading
Answers checked against the key, with question-level feedback.
Read the methodListening
What the learner understood, checked against accepted answers.
Read the methodThe evidence ledger
What we can prove.
Essay-level CEFR · ICNALE
94%
Within one band
37% exact. The band estimate has a margin of uncertainty; it is not an official exam result.
An excerpt from the Scale Trust Ledger. These writing findings do not establish speaking, reading or listening accuracy.
Read the full validation ledgerSource record
docs/reference/SCALE_TRUST.mdSay these out loud
The caveats belong
beside the claims.
CEFR is ±1 band at essay level.
The corpora themselves overlap that much. C1/C2 thresholds are provisional, pending the 500-essay top-end run after recalibration.
Absolute grades can be harsh in a classroom.
ELLIPSE’s best essays (about B2) map to US letter “D+”. That is correct on the absolute scale. Schools wanting classroom-relative letters can use per-school custom_boundaries.
TOEFL conversions diverge at the extremes.
The rubric-mean path and ETS’s IELTS↔TOEFL population concordance disagree at the low and high ends. The mid-range (B2) agrees.
The taxonomy count was corrected.
The public impact field skills_per_scan was corrected from 100 to 50. A list of skills is not evidence that every skill is measured in every submission.
A shared language for feedback
Name the skill.
Keep the evidence.
GRAMGrammarGRAM.AGREEMENT.SVALEXLexical ResourceLEX.CHOICE.PRECISIONDISCDiscourse ManagementDISC.COHESION.TRANSITIONMECHMechanicsMECH.PUNCT.COMMATASKTask AchievementTASK.CONTENT.RELEVANCECOMPComprehensionCOMP.GLOBAL.MAIN_IDEAPRONPronunciationPRON.PROSODY.STRESSThe source defines 50 skills across these domains, including strengths and receptive comprehension. A code is a label, not proof that a diagnosis is correct.
Explore the complete taxonomySource record
modules/evaluation/taxonomy_v1.yaml · version 3