Were you unable to attend Transform 2022? Check out all of the summit sessions in our on-demand library now! Watch here.
AI-powered search API platform provider Algolia is acquiring privately held vector-search vendor Search.io in a deal being formally announced today. Financial terms of the deal are not being publicly disclosed.
Algolia has developed its own proprietary technology that enables organizations to search internal resources and websites. To date, Algolia’s technology has used a keyword-based approach for search, which benefits from artificial intelligence (AI) to help improve relevance. Search.io has developed its own system as well, though unlike Algolia’s core system, it doesn’t rely on keyword relevancy. Rather what Search.io has built is a vector database–based engine that uses AI to convert content into numerical values, where relevancy can be determined based on proximity to the next nearest number.
With its acquisition of Search.io, the goal for Algolia is to enable an even more accurate approach to site search, using the power of AI. For example, instead of just a basic search using one or two keywords like “women’s clothes,” Bernadette Nixon, CEO of Algolia, told VentureBeat that a more natural way to search would be to specify what the user wants. So she said that if her sister’s son is getting married, she would want to use a search query like “killer outfit for the mother of the bride.”
“Consumers question whether keywords are the most effective way for them to search when they’re shopping,” Nixon said. “What people want is to be able to search as they think.”
Event
MetaBeat 2022
MetaBeat will bring together thought leaders to give guidance on how metaverse technology will transform the way all industries communicate and do business on October 4 in San Francisco, CA.
Register Here
How Search.io’s vector engine enables semantic search
Search.io has branded the technology it has developed as Neuralsearch, which provides AI-powered semantic search capabilities. At its core, it’s a vector database that enables highly relevant search queries to be executed.
“The reason that the vector database is so much more powerful than previous incarnations of how you deliver semantic search, for example, is because it has been trained on literally billions of documents,” Nixon said. “So the vector engine is therefore able to make the connections and give better context.”
Nixon explained that in the vector engine, the content is computed into a number that is multidimensional, meaning there are multiple associations with other things in the same index. She added that there are also computations in there as to distance from other things because that also affects and impacts context.
She noted that with vector engines, a concern can often be that it is more expensive to do the processing, storage and retrieval, due to the conversion of data into floating point numbers. That’s actually where Search.io has taken a unique approach with its Neuralsearch technology, which uses an innovative hashing technique to enable the vector engine to scale without needing specialized hardware and infrastructure.
Combining keyword and vector engines will enable a new type of site search and better recommendations
A traditional keyword-based search index is very different from a vector-based index. What Nixon said her team plans to do is bring to market a hybrid search engine that combines both keyword search and vector search.
Nixon said that Algolia will need to maintain two different indices, but that will be abstracted to users. Algolia’s technologies use an API that an organization can connect to in order to query and get search results. So what will happen with the new hybrid keyword/vector search will be that Algolia combines both of the indices into a single API call. As such, a user will make a query via the API that can then be sent to both engines, with a result that provides the highest level of accuracy and relevance.
Algolia has a range of technologies, including search engines, as well as a recommendation product that suggests products to users. The recommendation engine will also benefit from the Search.io technology that will bring in new AI models to help improve results there as well.
“Both companies have a long history of really focusing on relevancy,” Nixon said. “Combining the capabilities that we have as the two companies is what is going to be able to make us be able to have the most performance and the most cost effective results on the market.”
