Data extraction

A step-by-step guide to uploading papers, designing your extraction form, testing it, and checking the data and report against the source PDFs.

1. Choose Data Extraction

Open New Task and select the Data Extraction card. You can upload PDFs directly, without running a screening first. Abstractus will draft a form describing what to extract from each paper. You can review and edit the form before paying for a pilot or full run.

Select Data Extraction to upload your papers. The orange outline marks the control to use.

After you create the extraction, setup has four phases:

Documents
Form
Pilot
Full run

The examples below use fictional papers. Button counts and credit totals will reflect your own files.

2. Upload your PDFs

Drag your PDFs or a ZIP file into the upload area, or click it to choose files. Wait until the files are ready, then check the paper and page counts. Each extraction supports up to 80 papers and 1,000 pages. Each PDF can be up to 25 MB and 300 pages.

Click the upload area to choose PDFs, or drag files onto it.

Extraction costs 4 credits per page, with a 50-credit minimum per job. The pilot and full run share that minimum. Drafting the form does not charge credits.

3. Describe the information you need

What is this set of papers about? Enter the title to use for the report.
Your question: optionally add the research question you want the papers to answer.
What should we pull out of each paper? Describe the columns you would create yourself. Be specific about populations, outcomes, units and time points. Enter at least 20 characters.

For example: Extract the study design, participant ages, sample size, type and duration of CBT, comparator, anxiety measure, outcome at the end of treatment, and follow-up duration.

Select Draft the form for … PDFs. Once the draft is ready, review the documents and form in the setup wizard.

Select Draft the form after the uploads are ready and the required fields are complete.

Already completed a full-text screening?

Select Extract from a finished review instead on the upload screen, then choose Extract data beside your review. You can also select Extract data from the full-text results. That route lets you choose included PDFs and select Start extraction setup.

Only included papers whose PDFs are still stored are available. You must be the owner or an editor to start extraction. Starting extraction from a screening extends PDF availability by 14 days.

Use Extract from a finished review instead to reuse PDFs from a completed screening.

4. Review and edit the extraction form

In the Form phase, check the review title, question and unit of analysis. The unit of analysis describes what a record represents, such as one study or one study group. Open Fields to review each item to be extracted.

Field name: the label for the extracted value, such as Sample size.
What to extract: the information to look for, including the relevant units or time point.
How to judge a borderline case: explain which value to choose when several are reported and what to record if the information is missing.

Edit the field instructions and borderline-case guidance before running the extraction.

The form supports 10 to 30 fields. Use Add field to add an item, or the remove control beside a field to delete it. Set Study label field to the field containing the author and year.

Open Tables to choose the report's table titles, columns and the fields shown in each column. You can use up to 3 tables with up to 8 columns each. Open Protocol to review the extraction instructions.

For a broader revision, enter instructions under What should change? and select Redraft. The button shows how many drafts remain. Review the new form before continuing.

Select Save form to save your edits. If you move to the next phase with unsaved changes, Save and Continue saves them before proceeding.

5. Confirm the criteria stated in the report

The report criteria explain how the papers were selected. They are separate from the fields that tell Abstractus what to extract.

From a screening: review the restated screening criteria, edit them if needed, and select Approve.
From uploaded PDFs: enter any selection criteria and select Save. You can leave the text empty; the report will state that no criteria were given. You still need to save that choice.

You must confirm this section before continuing beyond the Form phase.

Select Approve for criteria carried over from a screening. For a direct upload, this button is Save.

6. Test the form on a few papers

In the Pilot phase, use the checkboxes to select up to 3 papers. If all three slots are selected, clear one checkbox before choosing another paper. Review the credit total, then select Run the pilot on … papers.

Run the pilot on the selected papers. The button shows the number of papers and the charge.

A pilot extracts only the selected papers and does not generate a report. When it finishes, review each field's value, supporting quote and status. If the results reveal a problem with the instructions, return to the Form phase and revise them.

If you keep the same form, pilot papers are reused in the full run and are not charged again. If you change the form, those papers must be extracted again and are charged again. An out of date badge identifies results from an older form.

The pilot is optional. Select Skip pilot · run all … to move to the full-run confirmation. If your pilot includes every paper, it costs the same as processing the whole set but still produces no report; continue to the full run to generate the report.

7. Confirm the full run

Continue to Full run and check the number of papers, fields per paper and remaining credit cost. Select Extract all … papers (… credits) to start. The full run processes the remaining papers and generates the report.

Select Extract all … papers after checking the remaining cost. This example reuses three pilot papers and charges for the other two.

The form, papers and pilot are locked while a run is in progress. You can follow the job on your Dashboard and open it when results are ready.

8. Read the report and check its sources

The results page places the report beside the PDF viewer. Select a linked citation in the report to open its source paper. Where a citation is linked to a specific extracted value and its passage has been located, the viewer highlights that passage. A general citation opens the paper without identifying a specific value.

Select a citation to inspect its source in the PDF pane beside the report.

Values traced to the PDF: how many extracted values have a highlighted source passage.
Quotes found verbatim: how many supporting quotes matched the paper's text exactly.
Citations flagged: report citations that need closer review. Check the cited paper before relying on the claim.

These checks describe the available evidence. A value may have a quote without a precise highlight, and a highlight does not replace reviewing the source.

9. Inspect individual values

Open Extraction matrix to compare papers. The on-screen matrix has fields in rows and papers in columns. Use Filter fields and values to narrow the display, then select a cell to inspect its source.

Select a matrix cell to open the evidence for that value.

Open Fields for this paper for the selected paper's values and evidence badges. Use Show p.… to jump to a highlighted passage, or Open p.… when only the page is known.

Select Show p.… to see a value's supporting passage.

Located: the passage has been identified and highlighted.
Approximate or page only: the match is partial, or only the source page is known.
Quote unverified: the supporting quote was not found verbatim. Check the source carefully.
Not reported or not located: the paper does not report the field, or its supporting passage could not be located.
Calculated: the value was calculated or read from a figure and may not appear as selectable text in the PDF.

If highlights need to be rebuilt, the owner or an editor can select Recompute highlights while the PDFs are available.

10. Export your work

Select Export at the top of the results page and choose a format:

Report HTML: download the report as a web document.
Report PDF: download a PDF version of the report.
Matrix CSV: download the extracted data with one row per paper for use in a spreadsheet.

Open Export and choose Report HTML, Report PDF or Matrix CSV.

Reports and extracted values are retained permanently. Original PDFs are retained for 14 days from the run. After that, the data and report remain available, but you can no longer inspect the original PDFs in the viewer.