Table of Contents
- ASReview Quick Tour
This tutorial assumes you have already installed Python and ASReview. If this is not the case, please check out the installation page.
Now, to launch the ASReview user interface, run the following in your shell (open the CMD):
In five steps you can start screening for relevant papers:
In Step 1, you need to create (or open) a project. Upon launching ASReview you arrive at the Projects page. Start a new project by clicking the red + sign in the bottom right corner, or continue with an already exisiting project by clicking the Open button:
Next, provide information on your systematic review project:
In step 2, select the data set you want to review, which should contain at least the titles and abstracts of all publications you want to screen.
There are four ways to select a data set: 1) upload your own data set, 2) import a data set from an URL, 3) select a plugged-in dataset (up until now offering the CORD-19 data), and 4) choose one of the built-in example data sets:
At step 3 you are asked to select some relevant papers: the publications of which you already know they are relevant for your systematic review. You can search your data set by authors, keywords and title, or a combination thereof.
Providing the software with prior information gives the software a head start. Note that there are no restrictions on the number of publications you need to provide, but preferably provide 1-5 prior inclusions.
Enter your search terms (for example “social”) and confirm by clicking the magnifying glass icon.
From the obtained search result, select the publication(s) you had in mind by clicking the heart icon. Click return to go back to the search engine.
Repeat this step until you’ve selected your 1-5 prior inclusions. Your prior inclusion(s) will be displayed below the search field and click the Next button.
Fourth, to train the machine learning model you also need to provide some irrelevant papers. This provides the software with additional information on what kind of publications should be excluded from your systematic review.
Given that the majority of publications in the data set is irrelevant, the publications presented here will most probable be irrelevant for your systematic review.
Indicate for each publication whether it is relevant or irrelevant to your systematic review and click the Next button.
In the final step you are allowed to choose a machine learning model. The default model is Naive Bayes, but you can opt for a different model if you would like to. After choosing your model, click Start Reviewing. Based on the information you have provided in Steps 3 and 4, the software is now building a machine learning model that predicts the next abstract most likely to be relevant.
If you have
.asreview project files, you can add them by clicking the red + sign in the bottom right corner.
Choose the project file from your computer and click Import.
Close the dialog and open the project to continue reviewing.
As soon as the machine learning model has converged, the software presents you with the publication of which the machine learning model is most confident that it should be included in your systematic review. You are asked to provide a label: relevant or irrelevant for your systematic search?
While you review the publications that the software presents you with, the software continuously improves its understanding of your decisions, constantly updating the underlying model.
The Statistics panel shows information on the current project and labeling progress. The panel can be opened and closed during labeling as you like. To hide your progress, click the closing arrow in the upper right corner.
The software keeps presenting you with the most relevant publication until there are no more publications left to review, or until you decide to stop reviewing.
As you keep reviewing abstracts, your set of relevant papers will increase while the number of unlabelled abstracts left in the data set will decline.
Now, in a ‘traditional’ systematic review, you would continue reviewing until you have seen all abstracts in the data set. However, ASReview orders the publications in such a way that you see the most relevant publications first. In other words, there is a point where you can be confident that you have seen (almost) all publications you need, and you can stop reviewing. When to stop is left to the user. A stopping criterium could be stopping after the last X presented abstracts were labelled irrelevant, or if your time is up.
Open the projects panel by clicking the 3-striped icon in the upper left corner. Click the export button.
Choose from the menu whether you would like to export your results as a CSV or an Excel file and click Export. A file is downloaded with the results of your review.
You can export your project as a
.asreview file by clicking Export below Download your project.
A file is downloaded with your project.