Fig 1.
The Flow Diagram illustrates the steps during document collection and evaluation. We skimmed more than 1553 papers and finally we selected a subset of 81.
Table 1.
Queries performed with advanced search for each database and the number of papers retrieved.
Fig 2.
The bar chart illustrates the trend of the selected publications over time.
Fig 3.
The pie chart shows the main publishers for the selected papers.
Table 2.
Examples of statements for specific biases and hyperpartisan statements for that bias.
Table 3.
Definitions of hyperpartisanship given in the selected papers.
Table 4.
This table describes the traditional Machine Learning algorithms used in the selected literature.
Table 5.
Features used with the best models described in Table 4. The features described in the columns are the following: Morpho-syntactic (MS), Lexicon (L), Semantic (S), Sentiment (SE) and Metadata (M). The approaches are: Style-based (SB) and Topic-based (TB).
Table 6.
Collection of the most performant deep learning models used in the literature.
Table 7.
This table describes the best performances of the models.
Table 8.
This table describes the datasets found in the literature.
Fig 4.
Language distribution in the datasets described in Table 8.
Table 9.
Table summarizing the papers selected with PRISMA methodology.