Semantic Search
What is semantic search?
Semantic search is the ability to find information based on what it means as opposed to matching keywords (keyword search). The example on the left highlights how semantic search can be more powerful over basic keywords. "Who leads this project?" returns information about the project manager despite not containing the any of keywords of the question.
Semantic Search
Searches based on meaningfully similar words and phrases.
Results are automatically ranked on how closely they match original search terms.
Returns relevant results beyond the original query.
Traditional Search
The user needs to type in the correct search keywords.
Requires extra algorithms to rank and organise results without checking meaning.
Often is constrained to the original search terms.
Neural Language Understanding
Natural language underdstanding (NLU) has been rapidly evolving over last decade thanks to advances in deep neural networks. Semafind leverages the state-of-the-art large language models to help you store and organise your knowledge. Behind the scenes, neural networks create a semantic index of your knowledge base.
Semantic Index
Each knowledge base is organised into a vector based representation using the output of a deep neural network.
The vectors contain information about what the sentences mean. Semafind then uses the similarity between the vectors to perform semantic search.
Can I use my own neural networks?
Absolutely, SemaDB is a vector database and on its own it doesn't run neural networks. You can and are encouraged to use your own neural networks. The Knotes Platform currently doesn't support custom neural networks.