Explore semantic search use cases
With semantic search, applications can provide a search experience built on context and intent, not just keywords. This type of search is helpful for many use cases. Here are some examples.
Semantic search use cases
Personalization
User preferences and activity can be captured as an embedding vector. This vector can be used to influence search rankings. For example, book recommendation systems might rank books by their similarity to a search query and a preference for historical nonfiction.
Knowledge management systems
The content on intranets and other knowledge management systems is often built over time with loose or unmaintained structures. Semantic search helps companies organize and find information based on the intent and context of documents, not just their keywords. This organization can be even more precise with an embedding model trained in the company's domain.
E-commerce
Semantic search allows applications to provide customers with relevant product results without relying on keyword matches. This method reduces the effort to maintain keywords or have awkward descriptions that optimize for lexical search. Instead, customers search by intent and meaning. This search type can also bridge the gap between technical domains, such as computer parts and customer vocabulary. For example, searching for "main chip" could match CPUs before other kinds of chips.
Nontextual use cases
There are many ways to use semantic search beyond text. The core feature of semantic search is computing the similarity of embedding vectors. A model can generate embeddings for text input or other inputs such as image pixels. An image model can be trained on object recognition to allow users to search for photos containing objects in a query photo.