4 ms·
Why cant facebook just go after blogspam clickbait in general? If a story is a copy of a copy, have the algorithm favor the original instead of the blogspam.
by basch 10y ago
Why cant facebook just go after blogspam clickbait in general? If a story is a copy of a copy, have the algorithm favor the original instead of the blogspam. If a local news agency copies an AP/Reuters story word for word, have the software suggest replacing the replica with the original. Treat local news duplicates the same as any wordpress+adsense blog. If journalists are adding commentary thats a different story (albeit most commentary is a needless act of faux value addition anyway) but mashable et all are just embedding ingur/youtube content with a catchy headline and the letters h/t appended to the end. This is a continuation of what ebaumsworld did since foreverago. Deaggregate the aggregating middlemen.
Ripping the embed out of the blogspam, and turning it into a direct link, is trivial. If the only content on a page is a youtube video covered in a hundred ads, just return me the youtube video. Flag websites as blogspam, and more agressivly return just the content they "stole." Make it hard to get off that shitlist. Id rather facebook make a little money deconstructing the web, than whomever spun up wordpress and paid for a bad theme being rewarded for polluting it. Facebook has an opportunity to add through reduction; to add value by finding signal in noise, leaving noise on the cutting room floor. Deduplication should be something incredibly easy to do algorithmically, compared to fact checking. (Hell facebook is already half way there with trending, anthough it too often lumps semi-unrelated events.) Google News and Techmeme do a great job of it. Techmeme eloquently spells out the solution http://news.techmeme.com/111031/techmeme-revealed http://news.techmeme.com/111031/techmeme-revealed
Too many different ideas are getting conflated here. Clean up the spam/dupe mess FIRST, and sifting the sensationalized salon/dailycaller hyperbole from the blatant false news becomes easier (because we then arent are drowning in nearly identical but different copies of articles.)
- itsdrewmiller 10y agoMaybe follow the link and evaluate the ad:content ratio or just the raw number of different ad services on the page. Seems like the collateral damage to such an approach would largely improve the overall web experience.
- basch 10y agoI do think it important to consolidate duplicates of stories, and credit the original copy. Maybe a feature (that users can disable if it fails) that automatically finds the source and replaces the post with the original. Maybe two versions of the newsfeed, the default "sterilized" and a hidden but accessible "raw"
- gm-conspiracy 10y agoIf the ads load quickly, from few domains, and work well in mobile, you can assume it is not a legit news site.
- wmf 10y agoI hate blogspam, but Facebook has to be seen to be taking direct action against fake news right now. They can't afford to take detours to fix other problems first.
- basch 10y agothey are one in the same problem http://www.phdcomics.com/comics.php?n=1174 http://www.phdcomics.com/comics.php?n=1174 quite honestly, "mostly fake but slightly rooted in truth blogspam" is 10000x the problem actual straight fake stories are. vitamins, detox, antioxidants, etc, etc, are a much larger pool of fake news than fabricated propaganda. people are making up outrageous stories for ad revenue, not to win elections. state sponsored narratives are unlikely to get flagged by these types of measures because they will be published by credible sources. I doubt NYT will start failing snopes tests. I doubt every SPLC post will get flagged for being out of context or unverified.
- dabockster 10y ago> Why cant facebook just go after blogspam clickbait in general? Hard to do with just a few algorithms. They would need human eyes reviewing sources. And since Facebook seems keen on eliminating its human workforce through AI/ML, I doubt that they would suddenly embrace human reviewers again.
- basch 10y agoI entirely disagree. You extract the content, and diff it against other extracted content. You dont need a human eye to determine that NBC11 republished an AP/Reuters story word for word. This should be somewhat basic pattern matching. We are talking about a company that can brute force vanity onion addresses! Semantic analysis should be able to detect procedurally generated content farms. Human turking might be harder to detect, but once a site gets flagged, all its posts can be checked with stricter scrutiny. It is extremely easy (from a computational standpoint) to rip a youtube embed out of a page and directly link to the source. If mashable and yournaturaldietnews are known content embedders, more aggressively deconstruct their pages. There also has to be a way to crowd source content verification. They have 1.8 billion people, a subset of those people give good feedback. Maybe some peoples reports should be weighed more heavily than others, if they have a history of making valuable reports. Maybe youre thinking fake news is mostly political. A lot of it is DIY and Health Tips, tech lifehacks etc. Rehosted videos with a banner added on the top and bottom. Animals. Any content taken from elsewhere can be easily detected, similar to Tineye or reverse image searching.