Disinformation-NLP | study comparing the robustness of NLP based Disinformation

Disinformation-NLP is a Jupyter Notebook library. Disinformation-NLP has no bugs, it has no vulnerabilities, it has a Permissive License and it has low support. You can download it from GitHub.

This study uses a 6 x 7 x 4 design in terms of embeddings, models and test sets. Resulting in 42 NLP architectures, each tested on 4 noise1 levels. It is focussed around the 'MAIN' notebook. The primary 0 hypothesis was tested and rejected using simple t-tests as detailed in the accompanying paper, but further statistical analysis related to the secondary hypotheses (Friedman tests with Nemenyi post-hoc tests) can be found here. (Larger 'embedding dedicated' notebooks can be found in the 'Emb_DedicatedModels' folder. These notebooks use longer (less reduction) texts and the full unsampled datasets to obtain the most accuracte results if needed). Embeddings:.