SkyPhrase, the semantic search engine that understands what you say
Nick Cassimatis is a young pioneer, able to unite research into artificial intelligence and computer science with the latest advances in linguistics and psychology. His dream of mixing natural language with algorithmic processing was born early, when Nick was just fifteen years old. This is how he programmed an English-to-French translator that earned him being recognized in the Top 20 of high school students in the United States, according to the USA Today newspaper.
His ideas are based on what is known as the “Cognitive Substrate Hypothesis” (“Cognitive Substrate Hypothesis«). Cassimatis professionally aims to mix well-integrated data structures plus specific algorithms, in order to recreate artificial intelligence and language. To what end? Years after the invention of that first translator, this researcher, graduated in mathematics from the WITH and with a master’s degree in psychology from Stanford, and currently a professor at the Rensselaer Polytechnic Institute, he would develop a start-up with two students, Skyphrase. The objective of it is to apply artificial intelligence and its knowledge of human language in new semantic search engines that can perform personalized research. Cassimatis conceives human language as an extraordinary piece in that great puzzle that is intelligence, and sees it as a great learning tool, useful for technological advances:
Language is a good piece of that problem to work on because human language can do so much more than any other animal language, and because humans can talk about almost anything, it must be a really general capacity
The origins of Skyphrase
Prior to Cassimatis’s invention, some applications attempted to “teach” software vocabulary and grammar rules, as well as the use of statistics in performing searches. But to personalize them, we need to put the language in context. For example, if we want to know what “Obama” is, the results we get will vary between “President of the United States” and “Japanese city in Fukui prefecture.” Understanding how through natural language we are able, through our brain, to discern these meanings, is still a puzzle for programmers.
However, there are tools like Wolfram Alpha, which works in a different way than it does Google. In the first case, the response is processed from a structured database. In the second option, a list of documents or web pages is provided in which the answer we need can be located.
Another interesting application in the case of Apple it’s Siri. Based on the motto “Your wishes are orders”, Siri was born as a recognition protocol together with the iPhone 4S. A similar application, developed by a Spanish company and called Sherpa, also allows you to carry out transactions through its use. The great disadvantage presented by both Siri and the Spanish tool has to do with possible privacy conflicts, when transferring confidential information to the network.
But if something characterizes both platforms, it is that both seek to solve the same question: how to process natural language to answer questions or make recommendations to users who use them?
The traditional search is transformed
With the invention of Skyphrase, available on the web, iPhone app and Chrome browser extension, a new era in the development of the personalized internet search. Cassimatis sums it up by saying that “Siri tries to memorize the Library of Congress, while Skyphrase tries to assimilate the language in order to read and understand the Library.”
The application, novel in terms of semantic searches, bases its development on four considerations:
- Just search for whole words.
- Understands conjunctions, coordinate sentences, and noun phrases.
- It does not search for synonyms or related words.
- The launch of this tool seeks to establish alliances with possible interested groups (linguists, experts in artificial intelligence, etc.).
Skyphrase, unlike Siri, and due to the improvements implemented in its creation, is capable of detecting information from Gmail Y Twitter, from statements such as “tweets from NASA with pictures” or “emails that Jane sent me during the holidays containing pictures”. Phrases apparently not at all simple for the search engines we are used to. Undoubtedly, with the generation of these new tools that use semantic tricks, the use of the Internet will be increasingly similar to the functioning of our intelligence and our brain, in the development of something as complex and exciting as human language.
Images | Search Engine People Blog, Flickr