Update images in README and docs

README.md

@@ -11,11 +11,7 @@ The Active learning for Systematic Reviews (ASReview) project implements learnin
 researcher. This way of interactive training is known as
 [Active Learning](https://en.wikipedia.org/wiki/Active_learning_(machine_learning)).
 ASReview offers support for classical learning algorithms and
-state-of-the-art learning algorithms like neural networks. The following image
-gives an overview of the process.
-
-
-![ASReview Command Line Interface](https://github.com/asreview/asreview/blob/master/images/Figure_ASReview_Pipeline.png?raw=true)
+state-of-the-art learning algorithms like neural networks.
 
 ASReview software implements two different modes:
 
@@ -43,7 +39,7 @@ pip install asreview
 
 The ASReview team developed a user-friendly user interface to replace the old command line interface. The new interface is still under development but is already available for testing and training purposes. We expect to release the interface in the upcoming weeks officially. See the installation instructions below the image.
 
-![ASReview Command Line Interface](https://github.com/asreview/asreview/blob/master/images/ASReviewWebApp.png?raw=true)
+[![ASReview Command Line Interface](https://github.com/asreview/asreview/blob/master/images/ASReviewWebApp.png?raw=true)](https://asreview.readthedocs.io/en/latest/quicktour.html "ASReview Quick Tour")
 
 Install the candidate release of ASReview with the command below. 
 
@@ -59,7 +55,7 @@ asreview oracle
 
 ## Covid-19 plugin
 
-![Covid-19 Plugin](https://github.com/asreview/asreview/blob/master/images/intro-covid19-small.png?raw=true)
+[![Covid-19 Plugin](https://github.com/asreview/asreview/blob/master/images/intro-covid19-small.png?raw=true)](https://github.com/asreview/asreview-covid19 "ASReview against COVID-19")
 
 The ASReview team developed a plugin for researchers and doctors to facilitate the reading of literature on the Coronavirus. The plugin makes the [CORD-19](https://pages.semanticscholar.org/coronavirus-research) dataset available in the ASReview software. We also constructed a second database with studies published after December 1st 2019 to search for relevant papers published during the Covid-19 crisis. 
 

docs/source/activelearning.rst

@@ -1,12 +1,12 @@
-Active Learning
-~~~~~~~~~~~~~~~
+Active Learning  for systematic reviews
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
 Active learning denotes the scenario in which the reviewer is labeling references that are presented by a machine learning model [3,4]. The machine learning model learns from the reviewers labeling decisions and uses this knowledge in selecting the reference that will be presented to the reviewer next. In this way, the annotated dataset starts out small and iteratively grows in size [4]. In the case of Automated Systematic Review, this process is necessary to provide some initial classifications to the model, because the reviewer starts with a dataset without annotations.
 
 However, in the general sense the key idea behind active learning is that, if we allow the model to decide for itself which data it wants to learn from, its performance and accuracy may improve and it requires fewer training instances to do so [5,6]. Moreover, we increase the dataset's informativeness by having the reviewer annotate those references that are more informative to the model. It has been found that active learning over a smaller dataset consisting of more informative data brings forth a model that is able to generalize even better than a model that takes the traditional approach of iterating randomly through a provided dataset with annotations [1]. 
 
 
-.. figure:: ../images/activelearningreview.png
+.. figure:: ../images/Figure_ASReview_Pipeline.png
    :alt: Active Learning for Systematic Reviews
 
 Different types of active learning are proposed: certainty-based active learning and uncertainty-based active learning. Certainty-based active learning, also known as max sampling, is typically the most suited to the scenario in which we are dealing with highly imbalanced datasets and finding all relevant references as soon as possible is quintessential to success [4]. This is because with certainty-based active learning, the model presents those references to the reviewer which are most likely to be inclusions first, thereby expediting the process of finding all of the relevant references.