Google has begun to use artificial intelligence in what it calls "sentence compression algorithms" to serve in desktop search results. The idea is to help machines carry out what has traditionally been a simple task for humans. The task is the ability to automatically generate and summarize a more direct answer, rather than just a list of search results.
The sentence compression algorithm, which demonstrates the advancements of machines to understand and respond to natural language in human speech, went live Tuesday. It analyzes large amounts of text and then rewrites it as a short summary to quickly answer search engine queries, according to Wired. Some refer to this as a rich or featured snippet in search results. Perhaps it will also improve search engine optimization.
"You need to use neural networks — or at least that is the only way we have found to do it,” Google Research Product Manager David Orr told Wired. “We have to use all of the most advanced technology we have."
The machines learn how to extract answers from long strings of text and tons of data by learning from humans.
Orr explained to Wired how the company used old news stories to train the company's artificial Q&A brain. It allows the machines to begin to understand how headlines serve as short summaries of longer articles, but the efforts are "painstaking" and often demonstrate the limitations of this type of machine learning.
A group of PhD linguists called the Pygmalion team, headed by Linne Ha, support the project.