propaganda-ner-model

Model Card for Model ID

from transformers import AutoTokenizer, AutoModelForSequenceClassification

Load the tokenizer and model

tokenizer = AutoTokenizer.from_pretrained("fatemamojibor/propaganda-ner-model") model = AutoModelForSequenceClassification.from_pretrained("fatemamojibor/propaganda-ner-model")

Example input text

text = "إن زعيمنا وحده هو القادر على إنقاذ الأمة من الكارثة التي تسبب فيها الغرباء. ويتعين على كل الوطنيين الحقيقيين أن يتبعوا قراراته ويدعموها دون أدنى شك من أجل استعادة عظمة بلادنا."

Tokenize the text

inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True, max_length=512)

Perform forward pass to get model output

outputs = model(**inputs)

Get the predicted class

logits = outputs.logits predicted_class = logits.argmax(dim=-1).item() print(f"Predicted class: {predicted_class}")

Direct Use

<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->

Use directly: the label is encoded as: {'O': 0, 'Appeal_to_Values': 1, 'Loaded_Language': 2, 'Consequential_Oversimplification': 3, 'Causal_Oversimplification': 4, 'Questioning_the_Reputation': 5, 'Straw_Man': 6, 'Repetition': 7, 'Guilt_by_Association': 8, 'Appeal_to_Hypocrisy': 9, 'Conversation_Killer': 10, 'False_Dilemma-No_Choice': 11, 'Whataboutism': 12, 'Slogans': 13, 'Obfuscation-Vagueness-Confusion': 14, 'Name_Calling-Labeling': 15, 'Flag_Waving': 16, 'Doubt': 17, 'Appeal_to_Fear-Prejudice': 18, 'Exaggeration-Minimisation': 19, 'Red_Herring': 20, 'Appeal_to_Popularity': 21, 'Appeal_to_Authority': 22, 'Appeal_to_Time': 23} {0: 'O', 1: 'Appeal_to_Values', 2: 'Loaded_Language', 3: 'Consequential_Oversimplification', 4: 'Causal_Oversimplification', 5: 'Questioning_the_Reputation', 6: 'Straw_Man', 7: 'Repetition', 8: 'Guilt_by_Association', 9: 'Appeal_to_Hypocrisy', 10: 'Conversation_Killer', 11: 'False_Dilemma-No_Choice', 12: 'Whataboutism', 13: 'Slogans', 14: 'Obfuscation-Vagueness-Confusion', 15: 'Name_Calling-Labeling', 16: 'Flag_Waving', 17: 'Doubt', 18: 'Appeal_to_Fear-Prejudice', 19: 'Exaggeration-Minimisation', 20: 'Red_Herring', 21: 'Appeal_to_Popularity', 22: 'Appeal_to_Authority', 23: 'Appeal_to_Time'}

Training Data

https://github.com/rafiulbiswas/ArAIEval_ArabicNLP24_Task_1

Given a multigenre text snippet (a news paragraph or a tweet), the task is to detect the propaganda techniques used in the text together with the exact span(s) in which each propaganda technique appears. This a sequence tagging task.

Citation [optional]

<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->

BibTeX:

@inproceedings{biswas-etal-2024-mememind, title = "{M}eme{M}ind at {A}r{AIE}val Shared Task: Spotting Persuasive Spans in {A}rabic Text with Persuasion Techniques Identification", author = "Biswas, Md. Rafiul and Shah, Zubair and Zaghouani, Wajdi", booktitle = "Proceedings of The Second Arabic Natural Language Processing Conference", month = aug, year = "2024", address = "Bangkok, Thailand", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.arabicnlp-1.50", pages = "494--500", }

Model Card Contact

[More Information Needed]