Google has always entertained its users by offering several useful features for content writing upgrades and this time, it is a new algorithm.
The new algorithm generates “coherent” articles by taking yours and competitor’s web pages in the account. The algorithm can resolve a user’s query without sending them to another web page.
How Does New Algorithm Work?
It works by summarizing web content using elements that extract your content by removing all unnecessary parts. The algo is similar to those algorithms generating featured snippets.
These algorithms are called “extractive summaries” as they extract content from web pages.
These extractive summaries are like reducing original content to the most imperative sentences only.
Now, after this, the new algorithm uses Abstractive Summary, a paraphrasing algorithm.
It has a disadvantage and that is the summaries include fake facts.
Google uses both approaches and that are “extractive summaries” are used to extract important information from the web pages and then applying the “abstractive summaries” is used to paraphrase the content. This approach creates a new document on the basis of web content and making Google’s own version of Wikipedia.
What Google Says-
“We show that generating English Wikipedia articles can be approached as a multi-document summarization of source documents.”
This suggests that Google uses several web pages to collect information about a topic.
“We use extractive summarization to coarsely identify salient information…”
This shows Google’s algorithm reduces the web page content to the most important sentences.
The next step is to use:
“…a neural abstractive model to generate the article.”
From “extracting” the content Google moves to a “neureal abstractive model” to summarize all those extracts into organic content form, better say an article.
Google says the final articles can pass human examinations.
“We show that this model can generate fluent, coherent multi-sentence paragraphs… When given reference documents, we show it can extract relevant factual information as reflected in… human evaluations.”