Quality assurance stage

To ensure high-quality data extraction, Laser AI includes a Quality Assurance module that allows a second researcher to verify the data extracted by the first researcher.

During Quality Assurance, the second researcher can review all data extracted by the primary researcher. They have the ability to accept or reject entries, as well as add additional ones. All fields requiring verification are highlighted in yellow.

To review extracted data in the Quality Assurance:

1. Click on the field name (e.g. study design). The list of all value types will expand. Accept the value entered by clicking on the green tick to edit values.  If you don't want to accept all values extracted from each field, click on the red ‘x’ mark. 

2. Review all author-reported values - clicking on each citation, will take you to the PDF’s corresponding section containing the reported value. You can remove any of these by clicking on the grey delete button to the right of the citation. Verification of the citations from the PDF will ease the verification of the second field - extracted values 

3. Review values extracted by first Researcher - data extracted here may be presented as a vocabulary field or as a text field.In both cases, you can edit the field by selecting a new vocabulary value (or create a new one if necessary) or enter additional data manually.        

4. Optionally, you can edit the commentary section or add your own comment to the particular field. 

Check if fields has any connections that need to be verified - if so, check that the appropriate links are established between the validated field and other elements of the data extraction form. You can also add additional connections by selecting the field and choosing the 'Connect’ button

During the Quality assurance, remember to check not only the correctness of the extracted data but also for any missing values.

If the first Researcher has missed any data that should be added, you can add it to the form during the QA process. Click on the ‘plus’ button next to the targeted field and you will see the fields to be extracted.  As with primary data extraction, these fields can be vocabulary or text, according to the rules defined during the development of the data extraction form.

You can check all rejected AI model suggestions by clicking on the “Rejected” button at the bottom of the page, and restore suggestions if needed. 

Remember to send the verified studies by clicking ‘Send ready extractions’ button in the reference list, as in the previous stages. Only when you send your data extraction forms, data can be further processed by Manager (e.g. in the Data cleaning stage).

RELATED ARTICLES

1. Tasks distribution-quality assurance stage  
2. Performing data extraction - introduction  
3. Performing data extraction - start here to understand the process  
4. Completeness check  
5. Modification of extracted data on Reference list  
6. Cleaning extracted data