Gaming is a ~$200 billion global industry — larger than film and music combined. Mobile accounts for roughly half. Within mobile, puzzle games generate an estimated $10–15 billion annually. Word games are a substantial segment — likely $2–4 billion, though precise figures are elusive since most revenue comes from advertising rather than app store purchases.
Word games occupy a distinct niche: smaller than mega-categories like RPGs or match-3, but characterized by exceptional user engagement and stable retention. Players return daily.
Linguabase was built to power word games. It’s been in development since 2011 and currently powers “In Other Words”, a live iOS game where players navigate semantic space.
What game developers need:
Word validation (is this a real word?)
Related words for puzzles and hints
Balanced sense coverage (not just the obvious meaning)
Graded difficulty (strong connections are easy, oblique connections are hard)
Word families for morphological puzzles
What Linguabase provides:
400K production vocabulary (1.5M available)
~40 core associations per word, sense-balanced
Connection weights for difficulty tuning
478K word family groupings
Narrative definitions for in-game display
Content blocklists — hard block for offensive terms, soft block for suggestive terms you may want to exclude from automated puzzles but accept as user input
2. AI Model Evaluation
The problem: How do you evaluate whether an LLM properly covers word senses? You need ground truth — verified sense-balanced associations to test against.
How Linguabase helps:
Ground truth for benchmarking sense coverage
Test whether models underweight non-dominant senses
Measure how well models capture directional asymmetries
Testing LLM on "bank":
→ Does it generate riverbank associations, or only financial?
→ Does it know pool table banking? Aviation banking turns?
Linguabase provides verified sense-balanced ground truth.
3. Educational Technology
The problem: Teaching concepts in isolation doesn’t build understanding. Learners need to see how ideas connect.
How Linguabase helps:
Generate concept maps automatically
Show how new vocabulary relates to known words
Sense-separated clouds for vocabulary depth
Word families for morphology lessons
Student learning: "photosynthesis"
Linguabase provides:
- Core: sunlight, chlorophyll, plants, carbon dioxide, oxygen, glucose...
- Sense (biology): cellular, organelle, chloroplast, leaves...
- Sense (chemistry): reaction, energy, conversion...
- Word family: photosynthesize, photosynthetic...
Automatic concept map with weighted connections.
4. Productivity / Writing Tools
The problem: Thesauruses give flat synonym lists. Writers need to understand nuance.
How Linguabase helps:
Connection weights distinguish “close synonym” from “related concept”
Sense labels show which meaning each synonym applies to
Directional weights help with word choice
Writer searching: alternatives to "big"
Thesaurus: large, huge, enormous, vast, massive, considerable, substantial...
Linguabase (with sense labels):
- size_physical: large, huge, enormous, massive...
- size_importance: significant, major, considerable...
- size_age: grown, adult, elder...
- size_informal: whopping, ginormous, humongous...
Writer can pick the right register and meaning.
5. Semantic Search
The problem: Keyword search misses conceptual matches. Searching “vehicle” doesn’t find documents about “cars.”
How Linguabase helps:
Expand queries using association graph
Find semantically related concepts
Rank by connection strength (strongest expansions first)
Delivery Options
Linguabase data can be delivered as:
TSV/SQLite files — For embedding in apps
API access — For server-side integration
Custom exports — Filtered by vocabulary, connection strength, etc.
We can also provide consulting on graph operations (pathfinding, distance calculation) and integration.
Not Our Target Market
If you need:
Dictionary lookups for e-readers → Use Oxford
50+ languages from one vendor → Use Oxford
Audio pronunciations → Use Oxford
“Powered by Oxford” brand credibility → Use Oxford
We built Linguabase for exploration and games, not reference lookups.
