With the Survey table extractor, it is possible to extract survey tables from PDF papers. The information presented in those tables are used to add papers to the ORKG. In this tutorial, we explain how to use this functionality.

What are survey tables?

Survey tables present information found in other papers, they are often presented in survey or review articles. Because the tables present information in a structured way, they can be used to quickly generate structured descriptions of those papers. Survey tables also contain references to the papers that are reviewed, based on the reference a description is generated for the referenced paper. Two examples of survey tables (note the references):

In case a table doesn't contain references, you cannot use this tool to import the table

How to import survey tables?

1. Upload the PDF and select the table

First, upload your PDF file to the tool. Afterwards, select the region of the table you want to import. Click the Extract table button to extract the table from the PDF. Table extraction from PDFs is a complicated process, the extracted table might contain errors which should be fixed manually.

2. Format the extracted table

The extracted table should be formatted in such a way that it adheres to the following rules:

• First row labels. Only the first row should contain the header labels (transpose the table if necessary, reformat the table if necessary)
• One paper per row. Each row should contain a single paper. It is not possible to use multiple rows to describe the same paper. Merge information in the same cells if necessary (to click to do, right click on multiple cells, and click "Merge cell values")

3. Map header labels

The header labels should be mapped to ORKG properties. This means that for each header label, you have to select or create an ORKG property. Do this by double clicking the header label, and selecting a property. In case a property does not exist, click "Create property name".

4. Extract references

The final step is to extract the references used in the table. In this step, the related metadata for each reference is captured. This includes the title, authors and publication date. To extract the references, click the Extract references button. In the popup that opens, select the column in which the references are used. Also, you can select the references formatting/style.

When the reference extraction is performed, 7 new columns are added to the table:

• paper:title
  this is a required column, each row should contain a paper title. In case the automatic extraction wasn't able to extract all references automatically, please add the metadata yourself. That means, go to the original paper, and copy the reference information into the respective columns. Rows that don't have a paper title, won't be imported.
• paper:authors
  contains the author names. Separate the individual authors by a semicolon (;)
• paper:publication_month
  often cannot be automatically extracted. Add the months manually, or just remove the column.
• paper:publication_year
  contains the year of publication.
• paper:doi
  contains the paper DOIs.
• paper:research_field
  can be selected by double clicking on the cell. It is not possible to create a new research field. Tip: use the same research field for all rows to save time.
• contribution:research_problem
  is a textual description of that problem that the listed papers address. The problem normally consists of 2-3 words, but more words are allowed. Tip: use the same research problem for all rows to save time.

When all cells are filled out, click the Import data button to start the import.