Hybrid search refers to merging search technologies — but the tech used to create hybrid search engines is changing.
Here’s where it’s at today.
Traditionally, hybrid search referred to a mix of directory- and crawler-based search engines.
Now, the meaning is changing to refer to a combination of keyword-based search technologies and artificial intelligence (AI).
AI replaces the complex rules-based understanding of traditional on-site search, while neural hashing makes vector-based search as fast as keyword search.
It might sound a little complicated — but the result is effortless searching for users.
Today’s hybrid search technology makes it possible for a query like “keep cool in summer” on an electronic store’s website to yield instant results for fans and air conditioners.
Without hybrid search, results might be limited to products that contain words used in the query.
In addition to offering greater speed and relevancy, hybrid search is more accessible for businesses to implement on their own websites.
A company called Search.io is one organization charting a path toward a new future for hybrid search, and I recently had a chance to talk to CEO and co-founder Hamish Ogilvy.
He informed me about recent developments in this field and how his company’s new tool makes it easier for customers to conduct searches on business websites. That, in turn, can lead to businesses making more sales.
AI + Neural Hashing = Modern Hybrid Search
In search, AI replaces strict keyword matching with dense vectors encapsulating text meaning. It’s shown to be superior to keywords when it comes to relevance, but comes with the tradeoff of slower results.
Neural hashing, sometimes called ‘deep hashing,’ comes in to make vector-based search as fast as keyword search. It gets its name from its ability to use neural networks to hash vectors.
Neural hashes compare terms using mathematical expressions. They measure the difference between words and concepts, and assign meaning to those closer together.
What does this mean in practice?
“In practical terms … AI understanding of language can be easily deployed in search technology. Ironically for many queries, despite the massively improved relevance, it is actually also faster than keyword search.
Neural hashing gives 99% the performance of dense vector search while being over 100x faster and using a fraction of the space.”
Ogilvy showed me several examples of natural language queries with and without neural hashing to show how modern solutions are failing retailers and customers.
A query like “something to keep my beer cold” returns a single, irrelevant result on retailer Best Buy’s website. In comparison, the same query with neural hashes applied would return a results page full of beer fridges.
That could make the difference between a customer leaving empty-handed, and a business making a quick sale.
Hybrid Search – More Accessible Than Ever For Businesses?
Search.io is launching a tool called Neuralsearch, which reportedly combines the speed of traditional keyword search with the accuracy of vector-based search.
In simple terms, it allows websites to return results more like Google does, without employing an army of search engineers.
It eliminates the need for retailers to add synonyms to their search index, which takes up a lot of time that can be better spent elsewhere.
Neuralsearch is now in public beta, following a private test with select organizations. Businesses can add it to their sites for free with a 14-day trial.
For an example of how it works in action, Ogilvy tells me the following company’s websites are already using Neuralsearch:
An earlier version of this story alluded to a shifting meaning of hybrid search as a combination of artificial intelligence (AI) and neural hashing. It has been updated to more accurately describe hybrid search as a combination of keyword-based search technologies and AI.
Featured Image: LookerStudio/Shutterstock