Auto-extracting Arabic lexical knowledge from machine-readable dictionary
Researchers used n-grams and rule-based IE to mine Al-Mawrid dictionary...
A team of researchers led by Diaa M. Fayed from Cairo University has developed an automatic method to extract linguistic knowledge from the Arabic-English Al-Mawrid machine-readable dictionary. Their approach combines n-gram analysis and keyword-in-context (KWIC) techniques to discover lexical patterns that reveal morphologic, syntactic, or semantic information. They then apply hand-crafted rule-based information extraction to pull out that data, also using punctuation marks and heuristics to identify synonyms within subentries.
The study, presented at the 5th International Conference on Arabic Language Processing (CITALA 2014) and archived on arXiv, focused on the Al-Mawrid dictionary. Results showed high precision across all information types, high recall for synonyms, but low recall for other extracted information. The work highlights that Al-Mawrid contains a significant amount of derivations (morphologic info), synonyms, domain labels, and hyponym/hypernym relations (semantic info). This automated extraction helps overcome the knowledge acquisition bottleneck for Arabic NLP applications, which require large lexical resources.
- Method uses n-gram and KWIC analysis plus hand-crafted rules for extraction
- High precision achieved across all extracted information types (morphologic, syntactic, semantic)
- High recall for synonyms but low recall for other information types
Why It Matters
Automates building Arabic lexical resources critical for NLP, reducing manual effort and enabling richer language understanding.