Hybrid search refers to merging search technologies – but the technology used to create hybrid search engines is changing.
This is where it is today.
Traditionally, hybrid search refers to a mix of directory and crawler-based search engines.
Now, the meaning is shifting to refer to the combination of artificial intelligence (AI) and neural hashing.
AI replaces the complex rule-based understanding of traditional on-site searches, while neural hashing makes vector-based searches as fast as keyword searches.
This may sound a bit complicated – but the result is easy search for users.
Today’s hybrid search technology makes it possible to make queries like “keep cool in summer” on an electronics store’s website, yielding instant results for fans and air conditioners.
Without a hybrid search, results may be limited to products that contain the words used in the query.
In addition to providing greater speed and relevance, hybrid search is easier for businesses to implement on their own websites.
A company called Search.io is blazing a trail into a new future for hybrid search, and I recently had the opportunity to speak with CEO and co-founder Hamish Ogilvy.
He briefed me on the latest developments in the field and how his company’s new tools are making it easier for customers to search on business websites. This, in turn, can lead to more sales by the business.
AI + Neural Hashing = Modern Hybrid Search
In searches, AI replaces strict keyword matches with dense vectors containing the meaning of the text. In terms of relevance, it proved to be superior to keywords, but at the cost of slower results.
Neural hashing, sometimes called “deep hashing”, is used to make vector-based searches as fast as keyword searches. Its name comes from its ability to hash vectors using neural networks.
Neural hashing compares terms using mathematical expressions. They measure differences between words and concepts and assign meanings to those who are closer.
What does this mean in practice?
Ogilvy explains:
“Actually… AI’s understanding of language can be easily deployed in search technology. Ironically, for many queries, it is also actually faster than keyword searches, despite being significantly more relevant .
Neural hashing provides 99% dense vector search performance 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 can fail retailers and customers.
On the website of retailer Best Buy, a query like “something that keeps my beer cool” returns an irrelevant result. In contrast, the same query with neural hashing applied would return a results page full of beer refrigerators.
This can make the difference between a customer leaving empty-handed and a business selling quickly.
Hybrid Search – Is Business More Accessible Than Ever?
Search.io is launching a tool called Neuralsearch that reportedly combines the speed of traditional keyword searches with the accuracy of vector-based searches.
Simply put, it allows sites to return more Google-like results without hiring a ton of search engineers.
It eliminates the need for retailers to add synonyms to their search index, which takes up a lot of time that could be better spent elsewhere.
Neuralsearch is now in public beta after private testing with select organizations. Businesses can add it to their website for free with a 14-day trial.
As an example of how it works, Ogilvy told me that the following companies’ websites are already using Neuralsearch:
- Unity.com
- BBC.com
- lockheed martin
- Sennheiser.com
- Catch.com.au
- RentPath.com
resource: search.io
Featured image: LookerStudio/Shutterstock
