Semantic Search: Technology That Understands What You're Looking For

Have you ever received completely irrelevant results when searching on a website - forcing you to scroll through page after page before finding what you actually need? This is where semantic search comes in - a technology that doesn’t just read the words you type but truly understands what you mean.

Traditional Search: A Struggle with Words

Imagine searching for "bass fishing." In a traditional lexical search, you’d likely get results that include both the word "bass" and "fishing," but it wouldn’t know if you’re talking about the fish or a musical instrument. For humans, the difference is clear, but for a computer? Not so much.

We collaborated with the Norwegian Directorate for Building Quality (DiBK) to improve their search experience. With an extensive amount of information on their website, traditional searches often resulted in imprecise matches. To help users find the right information faster, we took search technology to the next level.

From Lexical to Semantic Search

Lexical search works well for simple words or phrases, but when queries become more complex, we need smarter solutions. That’s why we implemented semantic search, which leverages AI to analyze text and understand the user's true intent.

How Are We Improving Search?

To enhance lexical search, we have:

• Implemented stop words (excluding non-essential words in searches)

• Adjusted Levenshtein distance to account for typos

• Prioritized key pages and documents

• Integrated a synonym dictionary (e.g., terrace and deck return the same results)

However, even with these improvements, traditional search has limitations - especially in large information archives.

Semantic Search: Technology That Understands Users

Semantic search goes beyond individual words. Using AI, we analyze both website content and user queries to deliver more relevant results by:

• Utilizing large language models to grasp textual meaning

• Vectorizing documents for better matches

• Enhancing question interpretation

For example, if you search for "How to build stairs?", the system understands that your query is about stairs - not just construction in general. As a result, building regulations for stairs will appear at the top of the search results, followed by related documents on railings and other construction requirements.

Lexical vs. Semantic Search: What's the Difference?

Semantic Search at DiBK

Together with DIBK we have launched an advanced search experience for technical building content.

We developed a dedicated platform for storing vectorized data and integrated it into DiBK’s website. Whenever an editor updates or adds new content, it is automatically indexed and made searchable. We use third-party AI tools to generate vector representations of both DiBK's content and user queries.

With the help of our skilled partners at Per Høj, we also introduced a sleek new homepage and search results design, enhancing not just the search functionality but also the overall user experience.

What Does This Mean for Users?

These improvements make work easier and more efficient for everyone dealing with building regulations, codes, and applications. DiBK’s new semantic search helps both professionals and private individuals find accurate information without wasting unnecessary time.

If you're tired of irrelevant search results and want to provide your users with smarter, faster, and more accurate information, we’re here to help. Let’s talk about how semantic search can transform your website and enhance your user experience.