Google is launching a powerful update to it’s search engine which uses semantic knowledge to understand entities and provide disambiguation. Google’s been developing it’s semantic capabilities under the hood over the last decade but with this upcoming search release it will become far more apparent to the end-user. The idea behind semantics is quite simple – rather than search for only keywords, the search engine actually understand some words so it can create context. If you search for “companies with many employees” you may find a published list which is titled “companies with over 10,000 employees”, “World’s 10 Biggest Employers ” or some variation of that. But if the search engine could understand that Microsoft, General Electric and Boing are companies then it could look for, and possibly aggregate, quotes like “Microsoft has 93,000 employees” and understand that Microsoft is a company. In fact, it may not even be necessary to go to the source document at all – Google could just present the answer.
Google calls this a “Knowledge Graph” which is built from indexed content and search behavior. It analyzes queries and the selected result (and many other factors) to improve its semantic understanding. Furthermore, it will provide disambiguation. Wikipedia users are very familiar with this – it gives you a choice of contexts to distinguish Mercury the element from Mercury the Planet. Google uses the disambiguation to significantly improve its understanding.
I’ve worked with Semantic Technology and it is very hard to develop algorithms which parse and understand human language. Google, however, has a real shot to get this right by leveraging its market scale through crowdsourcing – letting the user do the work. Ultimately this will further improve the accuracy of search targeting and market power.