Listen: Universal Query Classifier

We have developed a search query classifier that adapts to any label taxonomy instantly. Unlike traditional classifiers that are frozen to the labels they were trained on, this model lets you supply any list of labels at runtime. There is no retraining, ever. You just swap in new labels as your needs change.

Because the model treats labels as text rather than fixed category numbers, it can evaluate terms it has never seen before. It simply scores the semantic fit between a search query and your label text. This means you can roll out the exact same model across entirely different industries, from travel to legal services.

This flexibility is a game-changer for search engine optimization and paid search campaigns. You can map query intents at scale, analyze gaps on search engine results pages, or update campaign reports on the fly. As your marketing funnel evolves, you simply feed the model a new list of labels.

In performance testing, our large model achieved over ninety-one percent accuracy. It is also exceptionally well-calibrated, meaning it is highly reliable and rarely makes high-confidence mistakes.

Instead of being stuck with generic categories, you can now define what transactional or informational intent means for your specific business, and the model will follow.