Disinformation-Fake-News-Detection

Model Card for Model ID

<!-- Provide a quick summary of what the model is/does. -->

Fake News Detection Dashboard. The dashboard allows users to detect disinformation using six standard benchmark datasets.The front end, developed with Streamlit, provides an interactive interface for uploading files in CSV, PDF, or DOCX formats, and visualizes results using bar charts and word clouds. Its modular architecture separates data ingestion, preprocessing, model inference, and visualization, ensuring scalability and maintainability. The dashboard has been applied to multiple datasets, including EUvsDisinfo, EUvsISOT, EUvsIGF, FA-KES, George McIntire, and ISOT, enabling large-scale predictions, cross-dataset generalizability assessment, propagation analysis, and exploration of textual patterns contributing to disinformation.

Model Details

Model Description

<!-- Provide a longer summary of what this model is. -->

Inference uses cached pretrained models for efficiency. For ML pipelines, text is vectorised using stored TF-IDF unigram and bigram features before PA predic- tion. For DL pipelines, text is tokenised, padded, and passed through BiLSTM networks to generate probabilistic outputs, normalised to consistent binary labels (Fake/Disinformation or True). Results are displayed in real time with dynamic visualisations, including keyword-based explanations via TF-IDF for ML models and word clouds for DL models, ensuring transparency and interpretability within a unified modular framework.

• Developed by: Sadam Hussain
• Funded by [optional]: [More Information Needed]
• Shared by [optional]: Sadam Hussain
• Model type: Passive Aggressive and BiLSTM
• Language(s) (NLP): (NLP): English (en)
• License: MIT
• Finetuned from model [optional]: [More Information Needed]

Model Sources [optional]

<!-- Provide the basic links for the model. -->

• Repository: https://dashboard-fake-news-detection-xw2wv9dmgfsywawzpvn9wi.streamlit.app/
• Paper [optional]: https://doi.org/10.5281/zenodo.18158666
• Demo [optional]: https://dashboard-fake-news-detection-xw2wv9dmgfsywawzpvn9wi.streamlit.app/

Uses

<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->

Direct Use

The dashboard allows users to detect disinformation and fake news in English-language news articles. Users can input raw text, CSV, PDF, or DOCX files. Predictions are generated in real time using ML/DL pipelines and visualised through keyword explanations (TF-IDF) and word clouds.

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

[More Information Needed]

Downstream Use [optional]

1. Can be integrated into larger news monitoring systems or fact-checking pipelines.

2. Pretrained ML/DL/Transformer models can be fine-tuned for other datasets or domains.

<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->

[More Information Needed]

Out-of-Scope Use

1. Not suitable for languages other than English without retraining.

2. Should not be used to make legal, financial, or medical decisions without human oversight.

3. May produce incorrect predictions on highly domain-specific or adversarial content.

<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->

[More Information Needed]

Bias, Risks, and Limitations

1. Models may reflect biases present in the datasets (EUvsDisinfo, ISOT, FA-KES, etc.).
2. Model struggles to produce accurate predictions for short texts, particularly those containing 5–10 sentences or fewer than 100 words

<!-- This section is meant to convey both technical and sociotechnical limitations. -->

[More Information Needed]

Recommendations

1. Users should review model outputs critically.

2. Combine ML/DL predictions with human fact-checking for high-stakes decisions.

3. Consider retraining or finetuning if applying to new domains or languages.

<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

How to Get Started with the Model

Run each file DLtrain_models.py and MLtrain_models.py with python3 name of file.

[More Information Needed]

Training Details

Running Both File Available on Github repository \href{https://github.com/afridisadam1-alt/Dashboard-Fake-News-Detection/blob/main/DLtrain_models.py}{DLtrain-models.py} and \href{https://github.com/afridisadam1-alt/Dashboard-Fake-News-Detection/blob/main/MLtrain_models.py}{MLtrain-models.py}. After training the models for both ML and DL it will generate the pretrained models with vectorizers.

Training Data

1. Datasets: EUvsISOT, EUvsIGF, ISOT, FA-KES, George McIntire

2. Content: English-language news articles labeled as Fake/Disinformation or True

3. Preprocessing: Lowercasing, punctuation removal, tokenization; TF-IDF features extracted for ML models; sequences padded and tokenized for BiLSTM/Transformer models

<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->

[More Information Needed]

Training Procedure

<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->

Preprocessing [optional]

1. ML: TF-IDF unigram + bigram vectorization

2. DL: Tokenization, padding, batching

[More Information Needed]

Training Hyperparameters

1. ML: Standard scikit-learn implementations (Passive-Aggressive)

2. DL: BiLSTM, LSTM; batch size 32–64, learning rate 1e-3, early stopping

• Training regime: [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->

Speeds, Sizes, Times [optional]

<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->

[More Information Needed]

Evaluation

<!-- This section describes the evaluation protocols and provides the results. -->

Testing Data, Factors & Metrics

Testing Data

Held-out splits from all six benchmark datasets

<!-- This should link to a Dataset Card if possible. -->

[More Information Needed]

Factors

<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->

[More Information Needed]

Metrics

1. Accuracy, Precision, Recall, F1-score

2. Confusion matrices for interpretability

<!-- These are the evaluation metrics being used, ideally with a description of why. -->

[More Information Needed]

Results

Passive-Aggressive and BiLSTM models achieved F1-scores > 0.90 on most datasets [More Information Needed]

Summary

The system reliably detects fake news across multiple datasets, providing both interpretable visual outputs and real-time predictions.

Model Examination [optional]

1. Keyword-based explanations using TF-IDF (ML models)

2. Word clouds for DL models

3. Confusion matrices available for multi-class evaluations

4. LDA Topics and Themes based

<!-- Relevant interpretability work for the model goes here -->

[More Information Needed]

Environmental Impact

<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

• Hardware Type: NVIDIA GPUs (V100/3090) for DL/Transformer models

CPU for ML pipelines

• Hours used: [More Information Needed]
• Cloud Provider: [More Information Needed]
• Compute Region: [More Information Needed]
• Carbon Emitted: [More Information Needed]

Technical Specifications [optional]

Model Architecture and Objective

1. ML: Passive-Aggressive
2. DL: BiLSTM [More Information Needed]

Compute Infrastructure

[More Information Needed]

Hardware

1. NVIDIA GPUs (V100/3090) for DL/Transformer models

2. CPU for ML pipelines [More Information Needed]

Software

1. Python 3.10

2. PyTorch, TensorFlow, scikit-learn, Transformers, Streamlit, NumPy, Pandas [More Information Needed]

Citation [optional]

https://zenodo.org/records/18158666

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

BibTeX:

[More Information Needed]

APA:

[More Information Needed]

Glossary [optional]

<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->

[More Information Needed]

More Information [optional]

1. Dashboard supports CSV, PDF, DOCX file input

2. Visual analytics: label distributions, word clouds, keyword explanations [More Information Needed]

Model Card Authors [optional]

Sadam Hussain

[More Information Needed]

Model Card Contact

[email protected] [More Information Needed]