Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

> The extension also rates credibility of news articles you read.

So we can just meta-trust "The Factual" (and its what, neural net?) to tell us what's true? I don't see anything about transparency on the website. Granting that you have the best intentions to begin with, isn't that a massive moral hazard once you get popular?



https://www.thefactual.com/static/extfaq.html We try to be transparent. You should form your own opinion. We provide a tool. Use this tool as just another data pt.

Posting here so people don't have to click away:

* Why should I trust The Factual Credibility grade for articles?

The Factual automatically calculates the probability that an individual article is credible. This is based on four factors:

- The diversity and quality of its sources

- The factual tone of the article's writing style

- The expertise of the journalist on the topic based on their article history

- The site reputation based on historical scores of every article on the site

Because the calculation is automated and devoid of human involvement, criteria are consistently applied across articles and sources.

Also, because grades are specific to an article and not a publication, scores vary within a publication.

Our algo/tech can tell if something is good, rather than saying if something is surely bad. Check out our news site that we built using our rating system, if you have time: https://www.thefactual.com/news

More relevant to the thread : https://story.thefactual.com/news/story/229376-Mike-Bloomber...


"Because the calculation is automated and devoid of human involvement, criteria are consistently applied across articles and sources."

How is the training data generated?


For the tone of voice analysis, our training data was a large set of articles from Reuters and Associated Press - both wire services that are used by most news outlets. We supplemented this with a dictionary of emotional weights for words and some other heuristics.

For author expertise, we classify every article into one of 1000 different topics based on the IPTC taxonomy. We’ve now evaluated about 7M articles creating an expertise database on 50,000 journalists, i.e. what topics have they written on in the past and how much do they focus on that.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: