ORKG content, of individual contributions or contribution comparisons, can be used in downstream data science. Since ORKG data is highly structured, with the ORKG API and the ORKG Python library you can easily access and reuse ORKG content in data science and for your own purposes. Here we give you a quick intro and some examples to get you started.

Requirements

• We suggest using Jupyter here, but this works with a Python setup as well. If you have not installed Jupyter, please do so. If you don't want to or cannot install Jupyter on your local machine, you can use one of the many Cloud environments, such as the one provided by EGI, Azure, or Google.
• Install the ORKG Python library by running $ pip install orkg.
• You need to either reuse existing ORKG contribution comparisons or create new ones, possibly by first adding the required papers to the ORKG.

Usage

With the following code you can access ORKG contribution comparison data and automatically load data into a pandas DataFrame after which you can manipulate the data as you normally do with data frames.

from orkg import ORKG
orkg = ORKG(host='https://orkg.org', simcomp_host='https://orkg.org/simcomp')
df = orkg.contributions.compare_dataframe(comparison_id='[ORKG_COMPARISON_ID]')

For this to work, you need to substitute the [ORKG_COMPARISON_ID] accordingly.

Examples

We have created examples for numerous domains, which we briefly overview here. In general, we link the relevant Jupyter notebooks as related resources and computation outputs such as visualizations as related images in ORKG contribution comparisons. These resources are shown below the comparison table. We organize the Jupyter notebooks in a GitLab repository. Jupyter notebooks directly execute with mybinder so that you do not need to install anything on your local machine if you just want to take a quick look at these examples. Please understand that mybinder sets up dedicated VMs upon request of execution of these examples, which requires some time. While you wait, you can see the progress by looking at the mybinder log. You can of course clone the GitLab repository and, assuming you satify the requirements listed above, execute the examples on your local machine.

• For our comparisons on COVID-19, we have developed numerous examples that plot comparison data, e.g. classical time series for basic reproduction number, geospatial visualizations of modelled estimates for total number of cases in Chinese provinces, as well as examples that integrate data from multiple comparisons and third-party data to learn something new about the virus, for instance with the Microbe-scope.
• For the comparisons on Global Mean Sea Level rise projections, we have developped example notebooks that plot the seal level rise projections as published in the literature, see for instance the plot for the year 2100 and RCP8.5.
• As an example in Materials Science, we use a comparison on solubility parameters of certain compounds to reproduce static plots in the original paper (see Figure 13) and give users a dynamic experience with this data.

Related

• Blog post on data science with the ORKG