Read about semantic search and how it takes information retrieval to the next level and puts information at our fingertips.
My great grandparents were married for more than sixty years. They had a telepathy that comes from living together that long. They could finish each other’s sentences or anticipate when one wanted a coffee or lunch. An exchange could go like this.
‘Remember those figs?’ Grandma said.
‘You’re talking about that holiday to France in ’76 and it wasn’t figs, it was apples.’
‘That’s right! They were lovely. Do you remember that village?’
From the outside, it seems impossible conversation. Grandma was searching for a memory, but she had the fruit mixed up, but Grandpa was still able to know not only the fruit that she meant but that Grandma was actually trying to recall a beloved holiday. This kind of intuition comes from not only understanding the question but also being able to understand the intent and context of the question as well as the questioner. This used to be a very human capability but now, and more and more commonly, we expect computers to have the same intuition when we ask questions of them. This computational intuition comes from a technology called ‘semantic search’.
It is a broad term to describe a search that is able to determine the intent, concepts, meaning and context of the words a person is using for their search. Semantic search can also incorporate (or even infer) other elements such as location and search trends, providing results that vastly supersede text-based search, which is limited to results containing the search terms used. (E.g., it would never understand that Grandma was actually wanting to talk about a specific holiday over forty years ago.)
Historically, Google’s traditional search algorithms were so good that they eschewed the use of semantics. About a decade ago that confidence disappeared. The explosion of data and connectivity even took the search behemoth by surprise. Google did an incredible thing for a company that size. They made the fundamental shift toward incorporating semantic technology and they have been reaping the benefits of the scalability, flexibility and power of the approach. By now, every single letter in the ‘FAANG’s employs some form of semantics and semantic search. The approach is gaining traction across sectors and ‘traditional’ companies. There are very good reasons for this change.
Semantics relies on the meaning of the words rather than the word or even the language used. ‘Cat’, ‘kitty’, ‘mog’, ‘chat’, ‘gato’ and ‘katze’ are different words in different languages but what they refer to, the meaning of those words, is the same. If you are interested in ‘mogs’ but all the content refers to cats and kitties, you’ll get the search engine kiss of death ‘zero results returned’. Semantic search knows it’s not the string of letters you are interested in but those animals that some people keep as pets that have a tail, whiskers, four feet and go ‘meow’. This understanding of the meaning of words is powered by a ‘knowledge graph’. The knowledge graph is a description of things and their properties and the relationships between things that the search engine uses to understand the intent and the context of your search.
A simple example, which even some traditional search engines are capable of, is when you misspell your search term. It’s pretty easy for a machine to decide you meant ‘apple’ even though your finger slipped and you typed ‘appl’. Semantic search goes event farther by knowing, via the knowledge graph, that you are probably interested in the fruit or the company. If the user searches for ‘Apple annual reports’, it will know they are interested in the company, because the knowledge graph associates ‘annual reports’ with companies and it knows Apple is a company. If the user searched for ‘Apple farm’ or ‘Apple nutrition’, the knowledge graph would know to return results related to the fruit.
First and foremost, your customers, your employees, they find what they are looking for. In the days before semantic search, users had to guess like Ali Baba at the Thieves’ cave what combination of words would unlock the content they seek. Organisations spent vast amounts of time and money on SEO-consultants anticipating all the possible combinations a user might try. It was exhausting and expensive.
With semantic search, results aren’t limited to only knowing the meaning of the term. The knowledge graph knows the context of what you searched for. Just like grandpa knew grandma was wanting to talk about the holiday and not specifically about apples. The knowledge graph used by the Ontotext Platform is dynamic and, via its text analysis, constantly gaining more knowledge as more content is added. Besides content about apples, the knowledge graph can return results about specific types of apples, other fruit, nearby farms that grow apples.
With semantics, it is possible to personalise results without requiring unnecessary intervention. The knowledge graph is responding to search activity. It understands the context of the search as well as the user doing the search. Insights provided are more human-centric rather than content-centric. The search results are closer to how humans understand the world not limited to the zeros and ones of how a computer must understand it.
Semantic metadata is increasingly the language of the internet and world wide web, under the thin layer of content for human consumption are layers and layers of metadata. The nice side effect is that content can be created for use by humans. Doing things like keyword stuffing for SEO benefits to the disadvantage of readability is a thing of the past. By using semantic search, content is marked up with the metadata used by search engines, most notably by Google.
Just as grandpa knew grandma so well, with semantic search our computers can anticipate the information and insights we are seeking. The exponential growth of data means that the problems the tech giants faced when they switched to using semantics are challenging even modest-sized organisations today. It is why text analytics, semantic annotation and semantic search has been more and more common place.
Want to learn more and see some examples?
White Paper: Text Analytics for Enterprise Use
Use the power of text analytics for your enterprise