Leveraging natural language processing and artificial intelligence, Israeli-based startup Twiggle has interconnected bits of information from people, places and things so the technology can
understand the relationship between words in the next generation of search.
The three-year old startup in stealth mode built an ecommerce search engine for companies like Best Buy or
Amazon, along with others selling electronics. The search technology understands relationships between words like "light" and "weight." So when searching for "light laptops," for example, the queries
returns descriptive results such as the weight of the laptop, not the hue or shade of color. The same occurs when typing in words such as "quiet dishwasher." The engine computes the noise decimal
level and returns a list of products.
"It's not a trivial thing, because the natural language processing analyzes the context in real time and then ranks the products," said Amir Konigsberg,
CEO of Twiggle. The former Googler co-founded the company with CTO Adi Avidor, who worked on the Google Now project.
Konigsberg said to expect two large retailers to begin using the technology
within the next two months. And while he declined to reveal the names of those companies, he said the technology could integrate into voice-activated devices like Google Home, Amazon Echo, or even
text-based searches on Google Shopping to help consumers "find the most reasonably priced, popular vacuum cleaner that doesn't make too much noise."
Rather than requiring people to type or
voice queries so the search engine can understand the phrase, Konigsberg said the search engine can understand the query using whatever keywords that best suits the person entering the phrase. It
integrates into a Web site or mobile application.
The AI technology can cognitively make decisions on the type of results to return. Konigsberg gave this example: "I threw a hammer at the
window and it broke, but what broke, the window or the hammer?" Common sense for a human would suggest the window broke, but a search engine based on AI would need to learn that cognitive
reasoning.