Fill or Show: Cue Rules and a Previous-Word Table for Arabic Case Vowels on the Prague Arabic Dependency Treebank

Case vowels on the final letter of Arabic nouns mark the grammatical role of the word and are usually left unwritten. We propose a fill-or-show policy for these vowels in text entry and evaluate it as a simulation on treebank positions: the system writes the vowel when it is confident and shows the position to the writer otherwise. The policy combines a frequency table keyed on the previous word with five cue rules (free prepositions, plural-waw verbs, inna and anna, four nouns of annexation, kana and kanat), and one confidence threshold is chosen on development data. On the test split of the Prague Arabic Dependency Treebank (4,327 definite nouns with a single gold short final vowel), the rules alone fill 18.74% of these positions at 99.88% accuracy. Rules and table at the development-selected threshold fill 39.98% at 98.21% accuracy, against 87.11% for always writing kasra at the same positions, and leave 60.02% for the writer. At a stricter setting of 0.95 the policy fills 34.74% at 99.87%. If the writer answers every shown position correctly, accuracy over all 4,327 positions is 99.28% at the selected threshold and 99.95% at 0.95. Data and code: https://doi.org/10.5281/zenodo.24407151

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-21
DOI
https://doi.org/10.5281/zenodo.24414152
Primary Topic
Natural Language Processing Techniques
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Fill or Show: Cue Rules and a Previous-Word Table for Arabic Case Vowels on the Prague Arabic Dependency Treebank

Adel Amer
Zenodo (CERN European Organization for Nuclear Research)
Natural Language Processing Techniques
preprint

Fill or Show: Cue Rules and a Previous-Word Table for Arabic Case Vowels on the Prague Arabic Dependency Treebank

Adel Amer
preprint en

Abstract

Case vowels on the final letter of Arabic nouns mark the grammatical role of the word and are usually left unwritten. We propose a fill-or-show policy for these vowels in text entry and evaluate it as a simulation on treebank positions: the system writes the vowel when it is confident and shows the position to the writer otherwise. The policy combines a frequency table keyed on the previous word with five cue rules (free prepositions, plural-waw verbs, inna and anna, four nouns of annexation, kana and kanat), and one confidence threshold is chosen on development data. On the test split of the Prague Arabic Dependency Treebank (4,327 definite nouns with a single gold short final vowel), the rules alone fill 18.74% of these positions at 99.88% accuracy. Rules and table at the development-selected threshold fill 39.98% at 98.21% accuracy, against 87.11% for always writing kasra at the same positions, and leave 60.02% for the writer. At a stricter setting of 0.95 the policy fills 34.74% at 99.87%. If the writer answers every shown position correctly, accuracy over all 4,327 positions is 99.28% at the selected threshold and 99.95% at 0.95. Data and code: https://doi.org/10.5281/zenodo.24407151

Zenodo (CERN European Organization for Nuclear Research)
Natural Language Processing Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon