Latest labeled

Added Support for Python 3. Show latest labeled items so that it is easy for the users to change their decisions.

FAST2 data update

This version provides the updated datasets as .

The algorithm itself solves the following three problems:

• How to start
• When to stop
• How to correct human errors

FAST2 with human error correction

Apart from changes in FAST2, this version allows user to change their decision on previous labels. Machine suggestions used to efficiently handle human errors or concept drift.

FAST2

Core update with the techniques for 1. how to start and 2. when to stop

• starting review with either importing previous data or a keyword search is enabled and recommended;
• stopping review when there is few relevant papers left (according to the estimation).

Core Algorithm Update

Core algorithm is updated to utilize uncertainty sampling, Weighting, and aggressive undersampling.

Estimate number of relevant studies

Estimate the number of Relevant studies in the pool. Enabling Estimation will slow down the training, but provide the following benefits:

• number of Relevant studies will be estimated, thus helps to decide when to stop;
• probability scores will be more accurate;
• probability scores can then be used to decide whether to choose reuse or certain when Re-Using FASTREAD.

Apply old knowledge

Add UPDATE/REUSE features to allow FASTREAD import old knowledge from previously exported data to bootstrap a new review.

FASTREAD

Implement FASTREAD to support primary study selection in systematic literature review.