Google's Plan to "Estimate Web Sources' Trustworthiness" This short video is about more than Google's next plan to filter/bias search results. It is food for thought about the way people think and learn, and how that can be influenced by the way technology is applied. [yt]https://www.youtube.com/watch?v=be1W32qQy_k[/yt]
That entire video stems from a total misunderstanding of what Google is doing, and what Google (and computational linguists) is calling a "fact". To Google, a "fact" (or in my field, a factoid) is a verifiable basic piece of information about a subject. Examples would include: Barack Obama is male. Barack Obama is born in 1961. Barack Obama's wife is Michelle Obama. Barack Obama has two daughters. Robert Downey Junior plays the role of Tony Stark/Iron Man in Iron Man 1, 2, 3 and Avengers 1 & 2. Such factoids are irrefutable. If a website states that Barack Obama is female, then the website is in error. The video on the other hand, is talking about stuff that aren't factoids.
Intrinsical, could you elaborate on how Google's efforts—as described by you—would improve searches? Since human journalists (for example) can editorialize what is allegedly straight "just the facts, ma'am" news, how can a machine be programmed to tell the difference? Or are "factoids" only blandly obvious things, like "water is wet"? Or suppose a space news article states "scientists identify new black hole" (conclusion) when in reality all they actually observed was gamma rays and x-rays, does the filter mark it up or down? Please understand that I am not arguing any of the points above; I am interested in the technology Google is developing. It's not necessarily a "good thing." I read some apocryphal story that the first computer virus was unintentionally written as a self-replicating party invite that ended up gobbling computer resources. And the big technical gaffe in JURASSIC PARK (the book; never mentioned in the movie) was that the computer was told to sound an alarm if it came up short on the dinosaur census. No one ever imagined the computer needed a routine to say, "Hey, I just found more dinosaurs than quota." Small thing, but big consequences.
First of all, bear in mind that Google has spent the past twenty years improving upon its already excellent search results. Twenty years ago, it was only using one single factor to rank relevant pages: the number of links pointing to a webpage. Since then, they have been adding many more factors that can adjust a webpage's rank. I'm not working in Google, so I don't know what these factors are but I can make educated guesses. Factors could include the length of the page, the number of advertisements, how many times keywords appear in the page, how rare the keywords are, type of page (is it an article? a web form? a picture gallery? etc), is it a piracy-related page, is the language used in the page the same as the searcher's language, etc... From what I have read in research papers, Google now consider at least 200+ different factors. And each factor makes some small adjustment to the rank of each page. The use of accurate factoids would just be one more factor out of these 200+ factors. So as you can imagine the impact of 1 out of 200 factors is usually going to be very minimal unless the page in question contains a lot of erroneous factoids. As for what are factoids, they are like you said "really basic and blatantly obvious facts" about "named entities", primarily People (fictional, non-fictional, living, dead), Organizations (commercial, non-commercial, subsidiaries, departments, government agencies, clubs, societies, music bands, etc) and Locations (countries, states, towns, cities, streets, rivers, lakes, mountains, etc...). So for example factoids about people would include their name, birth date, family members, major affiliations, major accomplishments. People can include real life people as well as fictional people. So under this scheme we could even collect factoids about Jean Luc Picard, like he holds the rank of Captain, he's a member of Starfleet, he commands the USS Enterprise NCC-1701-D. For organizations like say Microsoft, factoids would include a list of Microsoft Directors, who is the CEO of Microsoft, the major divisions and subsidaries of Microsoft, Microsoft's annual revenue. For locations such as Senegal, factoids would basically be something akin to what can be found in the CIA World Factbook on Senegal. So its things like Senegal's GDP, land mass, population, etc...
Thanks, Intrinsical. Interesting stuff. It's obviously much more complicated than "Family Feud" where you're screwed if you're not "average."