When Pinecone introduced Vector Database early last year, it was building something specifically designed for machine learning and aimed at data scientists. The idea is that you can query this data in a format that the machines understand, making it much faster.
First, this includes semantic searches where users can search by definition rather than specific words. However, there were use cases where certain keywords were useful when people entered Pinco, and today the company announced that it is possible to search by combining definition and keyword searches, the company’s founder and CEO Edo Liberty hybrid search.
“We’ve done a lot of research on this topic, and in fact, we’ve found that hybrid search works best. [in many cases]. If you can combine both semantic search, this is a deep NLP sentence encoding that finds the context and the meaning and so on, but you can define it with specific keywords… It ends up being a combination of the two. To be significantly better,” Liberty told TechCrunch.
In fact, the two argue well, he says, especially in industry-specific terms. This could be something like a doctor looking for keywords related to a specific disease. In those cases, combining a query and specific keywords around a specific disease may return better results for the treatment scenario.
He said the keywords never come before the user’s semantic query, but they provide some additional information to return a more meaningful result.
“You might know exactly what you’re looking for, and you might be able to offer more offs when you let them know your keyword for semantic search — and that really helps a lot. So I don’t want to throw away the good parts of keyword research. [by relying completely on semantic search]. I don’t want the keywords to be in the driver’s seat, but I shouldn’t ignore them either.
As Liberty told us during the company’s $28 million Series A round last year, search has big benefits for the company.
“The primary use of vector databases is for searching, and searching in the broadest sense of the word. It’s searching through documents, but you can think of search in general as information retrieval, discovery, recommendations, anomaly detection, and so on,” he said at the time.
Pincon launched in 2019 and raised $38 million, according to Crunchbase.
