Embeddings as continuous spaces of meaning

Up until this point, we've known a lot about features, frequency, distance, and probability. All these approaches solve one important problem: we describe objects through explicitly defined features. A word is a set of letters. A document is a bag of words. A user is a table of numbers. Embeddings change the very way we think.

• Embedding as a vector of numbers (PHP)
• Case 1. Semantic search on vectors manually (pure PHP)
• Case 2. Why Euclidean distance is worse than cosine similarity
• Кейс 3. Кластеризация смыслов без ML-библиотек
• Кейс 4. Получение embeddings и semantic search (PHP + внешний embedding)
• Кейс 5. Semantic search через RubixML (k-NN поверх embeddings)
• Кейс 6. Поиск по событиям и логам (очень нужный кейс)
• Кейс 7. Почему embeddings лучше правил
• Кейс 8. Embeddings как интерфейс между человеком и системой