By Carsten Kraus
Developments in semantic search allow people to find what they want online quickly and easily with potentially huge implications for ecommerce.
The next big step in search engine functionality is here. Semantic search seeks to improve internet search accuracy by understanding a user’s intent and the contextual meaning of the terms they use to generate more relevant results.
The ability to type a query in your own words and have the search engine understand the inferences of the language has powerful implications for ecommerce. For example, at the moment there are no websites that could handle queries such as ‘show largest LCD TV under £5,000 with 2 USB ports’ or ‘show lightweight camera with compact zoom’. With semantic searching, however, this becomes possible as it allows the concept of ‘lightweight’ to be defined in context, as well as allowing users to select search criteria across a much wider field than is currently available.
What’s more, it can even interact with users to refine an initial search. For example, someone keying in ‘lightweight trousers for friend’s wedding’ may be asked anything from: “What is your friend’s age and status?” to: “Is the wedding in summer or winter?”.
We worked with Germany’s leading travel site, weg.de, on its recently launched full semantic search function, which allows browsers to find holidays simply by using a combination of vague ideas, such as “with 2 children over Christmas somewhere warm”. To make sense of this, the system behind weg.de draws from a geographic database, containing information such as a region’s climate, terrain and sites of interest – from museums to restaurants. This is then combined with dynamic information, such as travel time, number of travellers and so on. To handle this, we created a Probabilistic Inference Engine that enabled us to build up a cloud of possibilities and then quickly remove examples that are not relevant, allowing the search function to solve complex, real-world queries.
For example, someone could be searching for “Scotland New Year with beach”. While this is a valid request, a normal inference engine would reject it, because it would associate the term “beach” with someone looking for a warm holiday – and it is never going to be “warm” in Scotland over New Year. However, the person making the search could simply want a beach view.
One of the most powerful aspects of semantic search is that it is an evolving process. As the volume of search data builds up, the system will start to ‘learn’ customer behaviour and develop its own understanding of words. Through what we term Semantic Collaborative Filtering, the system would be able to make certain key associations. For example, it could learn that single people using the search term “Mallorca beach” like to travel to El Arenal, while those with families prefer other beaches. This is knowledge the software will be able to acquire automatically, so even if the system has never encountered a specific query before, it can still compute the best possible results.
To take full advantage of semantic search our searching habits will have to change. Research shows that, currently, people want to be able to get the right search results without having to type too much – with around 80 to 90% of searches being four words or under. However, we’ll need to move on from such primitive searching to entering free-form queries in our own words. This is where the power of semantic search lies, and with the technology for intelligent searching now well within our grasp, it shouldn’t be long before more people start making the change and reap the rewards through better results.
Carsten Kraus is chief executive of Europe’s leading online search specialist FACT-Finder.