Once you have shared datasets (manually created by Reusing Tidy Datasets or by Automating Dataset Updates), you can also consider creating code that depends on one or many pins to further process datasets or pin other objects like visualizations, models, and so on.
For instance, we could use the
worldnews pin to create a deep learning model on a daily schedule. One of the state-of-the-art language models is GPT-2, which we can also use from R through the gpt2 package.
Let’s first install the package and dependencies,
You can then retrieve the
worldnews pin, apply the GPT-2 text generation model, and pin the result in a new pin – which essentially creates a simple data processing pipeline:
You can preview this pipeline at beta.rstudioconnect.com/connect/#/apps/6565/access
You can also automate this process by reusing the techniquest presented in the Automate Dataset Updates use-case.