Persuasion Detection in Informational ads on Facebook

This repository contains data sources and code files for detecting persuasion in informational ads.

Classifiers :

Task 1 (classifier-1)

We conduct a set of experiments with different text classification models and identify the best one for classifying an advertisement as commercial or informational.

Task 2 (classifier-2)

We conduct a set of experiments with different classification models to identify the best one for classifying an advertisement according to 5 persuasion strategies : Authority bias, Scracity bias, Consensus, Moral Appeal, Positive Emotions and Negative Emotions.

Link to the classifiers: https://share.streamlit.io/nardjes-am/inf_vs_com_model/app.py

Datasets :

All the data that was used for training and validation is available in this repository on CSV files. Data used for classifier-1 is hand-labeled, while training data for classifier-2 is auto-labeled. However, for the second classifier, we add some samples for validation that could be used to verify how classifiers perfom on manually classified persuasion (not based on vocabulary necessarily).