evidoria

← Back to browse

Good practice Imported

AAEE — King Saud University's Automated Arabic Essay Evaluator for Saudi School Students

Saudi Arabia · Riyadh · See the Saudi Arabia profile · See the Riyadh profile

Evidence: Observational / pre–post Top 80% 33/100 · Ask Evidence Copilot about this practice

King Saud University researchers built AAEE, one of the first automated essay scorers for Arabic, modeled on Saudi schools' official rubric. Tested on 350 real student essays across eight topics, it reached 90% accuracy and a 0.756 correlation with human graders.

350 essays
Essay corpus size
8 topics
Essay topics covered
90 %
Scoring accuracy
0.756
Correlation with human graders
0.913
Correlation achieved by later Random Forest model (Alsaif et al. 2023)

Details

Promoter
King Saud University, Department of Computer Science, College of Computer & Information Sciences
Period
2019 (published study; research-stage system)
Keywords
natural language processing, Arabic-language assessment, academic research, K-12 education

Context

Researchers in the Department of Computer Science, College of Computer & Information Sciences at King Saud University built AAEE, one of the first systems to automatically score Arabic-language essays, addressing a gap left by automated essay scoring (AES) research that has mostly focused on English.

Objectives

Automatically score and give feedback on Arabic-language essays written by Saudi intermediate and secondary students, modeled on the official school rubric criteria of language proficiency, essay structure and content relevance.

Activities

AAEE combines Latent Semantic Analysis with Rhetorical Structure Theory to assess both lexical/grammatical quality and the coherence of an essay's argument structure; it was evaluated against a purpose-built corpus of 350 essays across eight topics, independently scored by human instructors as ground truth.

Results

AAEE reached 90% accuracy and a 0.756 correlation with human graders; a later 2023 study using the same corpus found newer machine-learning approaches, including Random Forest models, could reach a 0.913 correlation, surpassing AAEE's benchmark.

Conclusions

AAEE remains a research-stage system, validated on a compiled essay corpus rather than deployed in live Saudi classrooms or national exams, but it established an important early benchmark for Arabic automated essay scoring.

Implementation
Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)
Staffing & skills
Researchers in the Department of Computer Science, College of Computer & Information Sciences, King Saud University

Conditions for success

  • A purpose-built corpus of real student essays independently scored by human instructors as ground truth
  • Alignment of automated criteria with the official school grading rubric (language, structure, content relevance)
  • Combination of complementary NLP techniques (LSA plus Rhetorical Structure Theory) to capture both lexical and coherence quality

Common failure modes

  • Validated only on a compiled research corpus, not a live classroom or exam deployment
  • Newer machine-learning approaches have already surpassed its accuracy (per Alsaif et al., 2023)

Where it fits

Governance type
university research project
Scale
research-stage, single-institution study (350-essay corpus)
Income level
high-income (Saudi Arabia)

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful