ETS-based in-memory storage for WordNet data with fast lookups.

This module manages ETS tables for synsets, lemmas, and relations with multiple indexes for efficient queries. Uses lazy loading to minimize memory footprint and startup time.

Storage Strategy

ETS Tables

1. :wordnet_synsets_{lang} - Main synset storage

   • Key: synset_id
   • Value: Synset struct
   • Type: :set

2. :wordnet_lemmas_{lang} - Lemma storage

   • Key: {word, pos, synset_id}
   • Value: Lemma struct
   • Type: :bag (multiple lemmas per word)

3. :wordnet_word_index_{lang} - Word to synsets index

   • Key: {word, pos}
   • Value: [synset_ids]
   • Type: :bag

4. :wordnet_relations_{lang} - Relation storage

   • Key: {type, source_id}
   • Value: target_id
   • Type: :bag (multiple relations per source)

5. :wordnet_ili_index - Interlingual index (shared across languages)

   • Key: ili_id
   • Value: {lang, synset_id}
   • Type: :bag

Performance

• Synset lookup by ID: O(1)
• Lemmas by word: O(1)
• Relations by source: O(1)
• Memory: ~200MB for full OEWN, ~50MB for Spanish, ~40MB for Catalan

Example

# Initialize storage
Storage.init(:en)

# Store synsets
Storage.put_synset(synset, :en)

# Retrieve synsets
synset = Storage.get_synset(synset_id, :en)
synsets = Storage.get_synsets_for_word("dog", :noun, :en)