Der ultimative Guide zur semantischen Suche

Was es ist, wie es funktioniert und was es für Ihre Website bedeutet

Introduction

The amount of data in the world is growing. We now have access to millions upon millions of resources online – which is both a blessing and a curse.

On the upside, having huge amounts of data at our disposal gives us more options, and can help us make more informed decisions. But there is such a thing as too much data, especially when you have to wade through an ocean of information to find something specific. 

Ultimately, having so much data at our fingertips only helps us if we can search it efficiently. Without smart search tools, scouring for information online can be like searching for a needle in a digital haystack.   

Search engines like Google are constantly updating their algorithms to tackle this problem and improve online searches. But this challenge extends beyond general googling – it affects website searches as well.

Ever tried to find something through a website's search engine and got no results back? It's rough, isn't it? You're left either feebly trying different search terms until something comes up, or sifting through all the information on the site, hoping you'll stumble upon what you're looking for.

This is where semantic search can help. In this guide, you'll learn what it is, how it works, and what it means for your website.  

In a nutshell, semantics is the study of what words mean, how they relate to each other, and where they appear. It goes deeper than that, but the important word to remember is context.

A non-semantic search is bound by the limits of keywords. When you enter a search term – red shoes, for example – it will bring up all the results that feature the keyword red shoes. It knows what you're searching for, but it's not smart enough to understand why you're searching for it.

This becomes apparent when you try to put in a search term that could be interpreted in different ways, depending on the context. Let's say you search for 'apple'. You could be searching for:

  • The fruit
  • The multi-national corporation
  • Apple Martin, the daughter of Coldplay frontman Chris Martin

Non-semantic search won't know which one you meant. It will just show you all the results with the word 'apple' in them.

Semantic-based search, however, is smarter than that. It uses artificial intelligence (AI) and machine learning to approach language the same way a human would. It doesn't just think about the words you're typing in. It also considers why, where, when, and how you're searching for them. Using this extra context, the search engine can give you much more accurate and personalized results.

You know you're using semantic search when…

You search for a restaurant and get map directions

The search engine assumes that if you're searching for a shop, restaurant, or any business with a physical building attached to it, you'll also want to know how to get there. So, using your location data, it brings up the closest branches, along with their opening times and contact details. 

You search for 'childrn's books' and still get results

If it understands your meaning, semantic search doesn't care about typos. Instead of failing to recognize what you meant, it will suggest a correctly spelled alternative, along with the results for that term.

You can get the latest scores by searching for football or soccer  

Semantic search recognizes synonyms – different words that have the same meaning (e.g. football and soccer) – and broader terms for specific phrases (e.g. bird and pigeon). So, you don't have to be exact with your words when searching.     

How semantic search works

When you type something into a semantic search engine, it will factor in a list of contextual details, such as:

  • your location
  • your search history
  • word variations
  • synonyms
  • trends

Then it looks at how the words in your query relate to each other, combines that with all the contextual details it's found, and uses this information to make the results as personal and relevant to you as possible.

It's aiming to emulate the way you would chat with another human being. When you talk to someone, so many contextual elements play a part: where and when the conversation takes place, how much you know about the topic, how well you know the person you're talking to, and so on. These are the kind of things a semantic search engine considers, so it can make searches feel like a natural dialogue.   

There are a bunch of systems and methods behind semantic search that give it this extra level of intelligence. Some of these include:

Natural language processing  

It includes algorithms that allow a semantic search engine to understand queries phrased in human language. Several techniques are used to make search queries easier to process, including 'tokenization', where search terms get chopped up into 'tokens' (words, numbers, punctuation), and 'stemming', which removes suffixes ('ings', 'ions' and the like) from the ends of words.    

Fuzzy matching  

This is the process that returns results even if they don't exactly match your search term. This is how you still get suggestions from search terms with typos in them. It's a flexible system - you can apply different levels of 'fuzziness' to specific products depending on how strict or relaxed you want to be.  

Latent semantic indexing

A mathematical technique that analyzes connections between sets of data, instead of treating each word or sentence separately. In simple terms, it finds the hidden (latent) links between words (semantics) so you can sort information (indexing) more efficiently and make it easier to find.  

The story of semantic search so far

Want to hear about one of the earliest incarnations of semantic search? Ask Jeeves.

In case you weren't around in the late 90s, Ask Jeeves was a search engine with a human face – the titular butler Jeeves. You could ask him questions in everyday language and he'd dutifully find you some answers. Jeeves wasn't truly semantic (and wasn't always helpful), but the search engine was an early stab at the concept of natural language processing that's integral to semantic search today.   

Unable to compete with more popular search engines in the noughties, poor old Jeeves sank into obscurity and was shut down. But semantic search would soon make a big comeback, thanks to Google. 

 

Hummingbird started the semantic search buzz

In 2013, Google introduced its Hummingbird algorithm. Hummingbird was designed to better understand the intent of user's searches, drawing on the power of Google's Knowledge Graph (a ridiculously huge database that finds information related to a search from different sources).   

The update has made Google seem more human. Ask it a question in a conversational manner, as if you were asking a friend, and it will understand what you mean – then answer in kind. So, if you ask it “How old is Tom Hanks”, it doesn't just bring up a list of websites about Tom Hanks. It finds the answer (63, at the time of writing) and presents it to you on the search engine results page.

 

RankBrain made semantic search smarter  

Hummingbird was revolutionary, but it had a flaw: it still couldn't give accurate answers to questions it hadn't been asked yet.

That's why Google introduced AI into the equation in 2015 with its new update, RankBrain. When someone searches something new, the AI can bring up queries that have been asked before, find the closest match, and return results for it. It learns over time which results get clicked on most and uses that data to refine results for future searches.    

 

BERT is bringing even more intelligence to search 

In 2019, Google rolled out yet another algorithm update - Bidirectional Encoder Representations from Transformers, or to use its friendlier-sounding acronym, BERT.

BERT helps Google understand the nuances of conversational searches even better. Using bi-directional language modeling, it looks at what comes before and after a word in a sentence for added context. There are other clever tricks involved, like transformers and masked language modeling. But the long and short of it is that BERT is another step towards human-like intelligence for semantic search. 

Which brings us to the present day. Thanks to updates like Hummingbird, RankBrain, and BERT, fast and accurate search has become a normal part of our daily browsing experience. 

It's the new standard, which means people expect it wherever they go online. And not just from Google – from your website as well.

What semantic search means for your website

Getting on board with semantic search isn't about trend-chasing. If your site doesn't have it, it could harm your user experience.

 

Without semantic site search, you could lose out

Imagine a customer, Sarah, tries to find something specific on your site – let's say a size 10 blue top that's under €50 – and you've only got keyword-based search. The top she's looking for doesn't show up because her search term doesn't match a keyword in your database. And the results that do appear are irrelevant. There's a pink top, a blue dress, a blue top in the wrong size, a blue top that's €110, but not the result Sarah was originally looking for.

So, she plays a game of trial and error with the search engine. She types in different variations – blue top small, tops under €50, blue womenswear – but the search results come back empty. Eventually, she gets fed up, clicks off your website and heads back to Google to find another website. Hey presto, you've lost yourself a conversion.

Perhaps this wouldn't have happened ten years ago, before the rise of semantic-based search. But if you want to ensure today's Google-spoiled browsers stay on your site and complete a buyer's journey, you need search tools that can understand the intent of their queries. 

Bring semantic site search into the picture, and suddenly the above scenario changes to a much more positive one, for both you and your customers. You can look forward to:

 

Better user experience

Think about what makes a great UX when browsing a website:

  • A fast, friendly, and intuitive interface
  • An easy, frustration-free search experience
  • Relevant and uncluttered results

That's exactly what semantic site search achieves. Features like autocorrect and autocomplete make searching quick and simple. It can interpret natural language queries, so you don't have to play tiresome guessing games with the search bar just type what you'd ask out loud. And you get the most pertinent results because it can read the context of the search and understand your intent. 

Having semantic search also encourages people to stay on your website if they want to look for something else. They won't feel the urge to return to Google if they can get the same convenience using your search features.

 

Higher conversion rates

If your search engine delivers the relevant results first time, it can only be beneficial to your conversion rates.

This is a no-brainer: the faster and easier it is for people to track down what they're looking for on your site, the more likely they are to buy something. On average, e-commerce sites with a semantic search bar experience a mere 2% cart abandonment rate, compared to the 40% rate on sites with non-semantic search. 

If customers have a great experience from landing on your site to completing a purchase, they're more likely to stick around and carry on browsing. Then they'll find more products that catch their eye, even ones buried deep in your catalog. And before you know it, you're making even more sales. 

A quick, intuitive search experience will also bring customers back to your site to buy other products. They might recommend it to others as well. Then those people visit your website, make purchases, recommend the site to more people, and the conversions keep stacking up. 

 

More personalized results

A one-size-fits-all approach to customer service won't cut it anymore. Consumers increasingly want their experiences to be tailored to them. 90% find personalization appealing, and 80% are more likely to do business with a company if it offers a personalized experience.   

Semantic site search gives them that personal touch. For instance, if they've searched for sweatpants and running trainers, it can prioritize sportswear for future search results. Or if they're out and about with a mobile device, it can point them to a nearby store. These are the kind of details that will keep them coming back to your website. 

Personalization always comes with privacy concerns - but browsers have nothing to worry about. Any data collected through search is never personally identifiable and they have complete control over settings for tracking cookies. 

 

Potato, Potahto – Semantics and Localization

One of the interesting challenges of natural language processing is regional dialects and colloquialisms. American Midwestern users, for example, would call a fizzy soft drink “pop”, while people from the Northeast would call it “soda”. To offer a smooth experience for everyone, semantic search needs to account for these variations. 

What to do next

Setting up a semantic search engine on your website can be a complex undertaking. So, to make your life simpler, look for a solution that you can implement easily, without needing loads of coding skills. At the same time, your solution should give you the flexibility to customize it if you want to get a bit more hands-on.    

A custom site search engine such as Site Search 360 would be an ideal candidate. Its semantic search features ensure your users can find what they want, not just what they type in. You can integrate it out of the box – all you need to do is install a plugin or copy and paste a few lines of code. And you can customize it to your heart's content, with no coding experience needed.  

See how it works in action by starting your free trial today.

Auf der Jagd nach einer besseren Suche?

Teste Site Search 360 auf Deiner Domain. Mit der kostenlosen Trial kannst Du selber erfahren, wie einfach exzellente Site Search sein kann.