Google researchers propose a new method for ranking search results, prioritising facts rather than reputation
A team at Google has suggested a new model for search, which is based on measuring the accuracy of facts in a web page, rather than its reputation.
At the moment, the position of a website partly depends on the number of other websites linking to it, so that the more times a page is linked to, the further up it appears in results. The problem is that it leaves plenty of room for false information spreading.
A new fact-based ranking system would prevent things with little factual basis going viral on the internet, and improve the reliability of what we end up reading.
“We propose a new approach that relies on endogenous signals, namely, the correctness of factual information provided by the source,” the researchers say in their paper, which is published in the Cornell University Library.
"A source that has few false facts is considered to be trustworthy.”
Under the system, facts would be extracted from each source and used to construct a “Knowledge Vault”. Websites would be given a knowledge-based trust score according to how much their facts match up to the knowledge bases.
The team has already shown that the method can reliably work out the trustworthiness of different sources.