California-based semantic search experts at Cognition Technologies know their semantic search business and according to them, search engines like Google, Cuil and Yahoo! Search Monkey are not the real deal, when it comes to true semantic searching
. CMSWire had a chat with Cognition to get the scoop on what a real semantic search engine should do.
Is Google Semantic Enough?
Recently, we suggested that Google Suggest
was an early iteration of the semantic web. Dr. Kathleen Dahlgren, founder and CTO at Cognition Technologies
According to Dahlgren, the real semantic search
is all about words and understanding of their meanings. The so-called semantic search engines like Cuil
and Yahoo! Search Monkey
are just "statistically semantic."
Without understanding the words that people use to search, the search engines are not returning very relevant results. Dahlgren reports that Google, for example, has 20% precision
with its searches and only 20% recall.
Cognition Technologies boasts that their searches yield 90% relevance. Although, we must say, in Google's defense, that Google Suggest is not trying to focus on the linguistic nature of searches, but is merely gauging results based on the searches done by other people.
How the Real Semantic Search Works
According to Cognition Technologies, it's very rare for people to use the same sentence structure or the same words. Knowing that, the real semantic search technology should primarily focus on improving semantic technologies and learning new words, as well as their relationships with each other.
There are 506,000 word stems in the English language. An average college graduate has about 42,000 words at his disposal. Considering these two facts, the probability of repeating patterns is slim. Users' searches are limited by the confinements of search engines' rules and algorithms. Today's users of search engines are now trained to use one- or two-word search terms to help narrow down the relevant results.
Proper semantic search works in the completely opposite way. It relies on more that two words to make search results more relevant. The real semantic search uses the words most search engines regard as extra (such as "and," "but," "not," etc.) as a way to further extract the context, in which the words are being searched.
Understanding the Meaning of Words
No matter how many algorithms the big search engines boast, unless they are focused on understanding the meanings of the words being used, they are not really semantic.
If users really want to find what they search for faster and easier, they should consider companies with a narrow focus on the science of semantics and linguistics, like Cognition Technologies. Although the company doesn't spend a lot of time and resources on marketing, its technologies are being used by many prominent organizations in the legal and healthcare fields, including Medline and Lexis Nexis.