Abstract | ||
---|---|---|
Online propaganda is a growing concern. Fraudulent users write under multiple signatures to give the impression that the opinions they promote are more widespread than they really are, or held by a different demography. The problem as such is not new, but it is becoming increasingly organised and therefore has effects on a larger scale. In this work, we develop methods for assessing the true number of authors of a body of work, to detect artificially inflated user sets. The assessments are based on stylistic richness, here measured as the number of unique features (e.g., words or syntactic fragments) divided by the sum of all features. Initial results suggest that the order of magnitude can be reliable estimated. It is for example possible to differentiate the works of hundreds and thousands of writers. |
Year | DOI | Venue |
---|---|---|
2016 | 10.1109/EISIC.2016.48 | European Intelligence and Security Informatics Conference |
Field | DocType | ISSN |
Data science,World Wide Web,Impression,Computer science,Syntax | Conference | 2572-3723 |
Citations | PageRank | References |
0 | 0.34 | 0 |
Authors | ||
2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Johanna Björklund | 1 | 5 | 6.23 |
Niklas Zechner | 2 | 3 | 3.19 |