Antenor Challenge
Trademark conflict is not decided by spelling. Examiners and courts judge likelihood of confusion, and a mark that sounds like an existing one is a conflict even when it shares almost no letters with it. A plain text search over a registry will happily miss the collision between a mark written one way and a competitor pronounced the same way, and in Spanish and Portuguese the gap between orthography and pronunciation opens dozens of such traps.
For DeMarcas, whose business is trademark research across Latin America, that gap was the product problem: a clearance search that misses sound-alikes is a search a client cannot rely on.
Solution
Antenor built a phonetic search engine for trademarks. Instead of matching strings, the system encodes marks by how they are pronounced, using phonetic representations tuned to Spanish and Portuguese, and ranks candidate conflicts by similarity of sound. A query for a proposed mark returns the registered marks a consumer might confuse it with, not just the ones that happen to share characters.
The engine sits on ingestion pipelines that load and normalize trademark data from registries across Latin America, keeping the searchable corpus aligned with the official sources attorneys must answer to.
Stack
- Phonetic encoding tuned to Spanish and Portuguese pronunciation
- Similarity scoring and ranking of candidate conflicts
- Ingestion and normalization pipelines for Latin American registry data
- Search interface built for trademark professionals
Results
- Sound-alike conflicts surfaced that plain text search structurally misses
- Registry coverage across Latin America: Argentina, Venezuela, Chile, Uruguay, Brazil and other countries (currently Argentina)
- In production for DeMarcas’ trademark research practice
