March 24, 2017
The Dawn of Semantic Search in E-Commerce
Anyone who has ever wandered into a bookstore or library looking for a book they heard described on National Public Radio, or had recommended by a friend, will understand that often the book’s title escapes you. Which is why, when a well-informed bookseller or librarian is able to find the right book for you with the jumbled information you’re able to remember, you’re happy and impressed.
E-commerce sites are not quite there yet to provide a similar experience, not even a behemoth like Amazon.com. A search for “novel post-apocalyptic Michigan” on Amazon does not return a result that includes the correct book, “Station Eleven” by Emily St. John Mandel.
A knowledgeable bookseller or librarian would have known this title and so does Google: The top five search results are about this book and even include the book’s product detail page on Amazon.
Why is Google able to find this page and Amazon is not? The answer is: semantic search.
What is Semantic Search?
In short, semantic search improves search accuracy by understanding the contextual meaning of terms and therefore the searcher intent, comparable to the sales person in a brick and mortar store, who is able to understand the customer’s intent from tidbits of information.
Put another way, semantic search takes a human approach to data, by identifying relationships between topics, and beyond just keywords. While the Amazon listing for Station Eleven might not mention the keywords “post-apocalyptic”, semantic search would recognize the phrase “civilization as we know it came to an end” and identify the listing as relevant, in the same way as a human bookseller would.
Google added semantic search to its algorithm in 2013 and is currently taking it to an Artificial Intelligence level with RankBrain, which primarily aims to understand the intent of search queries that have never been seen before by Google. Interpreting natural language is key to this approach.
Semantic Search in E-Commerce
A retailer’s search engine that utilizes semantic search understands the natural language that a customer is using and presents the product that the customer was aiming to find, even if the customer and the retailer use different terms for the product. As a result, customers can more easily find what they are looking for and this match of search intent and result leads to an increase in sales and customer loyalty.
It is safe to assume that a leading e-commerce site like Amazon uses some sort of semantic search, but it doesn’t appear to be very good yet. Competitors who offer a superior site search experience might be able to better compete with the e-commerce giant.
Most likely, an outside vendor such as Twiggle, a company founded by two former Google employees, will develop this experience. What makes Twiggle’s solution especially attractive is that it doesn’t require online retailers to replace their enterprise search infrastructure, reported Search Engine Land. Their API can add a natural language layer to any search engine by passing each query in real time for an analysis similar to what Google does with its semantic search component.
“When you teach a search engine to understand people the way people understand each other – you take the guess work out of matching the right products to the right queries – and you can get best-in-class results,” stated Twiggle CEO Amir Konigsberg in a recent press release. “The kind of understanding (and customer service) that is effortless for a human store clerk can now be mimicked by a search engine.”
E-retailers utilizing advanced semantic search or even Google-level Artificial Intelligence in their site search will gain a significant competitive advantage, because happy customers are customers who return. This hasn’t changed in hundreds of years, and will not change with retail shifting from brick-and-mortar to e-commerce.