Below are some answers to some of our frequently asked questions.
Is News Bot really a bot?
Yes. News Bot is an actual bot which should be evident by its frequent posting to multiple platforms.
It's automated to post approximately every hour, however posting times may differ as some content takes longer to generated and errors might delay things as well.
How is News Bot so coherent?
News Bot was made in a way that made it "coherent" from day 1. It used pre-made sentence formats in which keywords just had to be put into.
This helped shape the initial dataset for the bot.
Keywords are also mapped using relations and was put together by thousands of people using News Bot's submission feature.
How are images retrieved?
Images are retrieved from Unsplash using their API.
The image for an article is based on its keywords and categories but there is no guarantee that an image fits perfectly as it's randomly chosen.
What technologies does News Bot use?
News Bot uses many different technologies such as:
Programming Languages: C#, Python and D
Libraries / Frameworks: .NET, GPT-2, Tensorflow, vibe.d, Mir
How are headlines created?
Headlines are created using a few different algorithms including but not limited to:
Generic sentence parsing
Generic sentence formatting
Keyword mapping by relations
Mapping keywords to generic sentences
Data comparison to previous generated data
How are articles created?
Articles are created primarily using GPT-2 using headlines as conditions.
What model did you use and how was it trained?
The 355M model from GPT-2 was used and trained using actual news both real and satirical.
Will You Open Source Your Code?
There are plans to open-source part of the code but not everything.
We believe it's better to keep most of it a secret as the code can be abused and used to create deceiving news articles, headlines etc.
Remember News Bot Lab is not creating "fake news", we're creating fictional satirical news.
None of the articles should be taken as fact. We're not trying to deceive.
However the technology we use could be used for deceiving and that's why parts of it will always remain closed source. To protect the general public.
How are categories attached to articles?
The initial dataset was manually categorized and afterwards all articles uses that as reference on how to attach categories to themselves.
This means the category data is constantly being updated automatically and becomes more fluent as time goes by.
Are the editors, journalists or reporters real?
No, the faces are generaed using https://thispersondoesnotexist.com/ and the names are references to different things such as technology and science.
If you have anymore questions, feel free to contact us
Want to advertise your brand, product or website here?
Click here to find out how.