README.md

---
language: en
license: mit
datasets:
- Twitter15
- Twitter16
- PHEME
tags:
- conversational
inference: false
model-index:
- name: Rumour Detection
  results:
  - task:
      type: conversational
      name: Rumour-Detection
    dataset:
      name: Twitter15, Twitter16, PHEME (Retrained scores)
      type: train and evaluation dataset
    metrics:
    - name: F1
      type: f1
      value: 0.698
  - task:
      type: conversational
      name: Rumour-Detection
    dataset:
      name: Twitter15, Twitter16, PHEME (Scores reported in paper)
      type: train and evaluation dataset
    metrics:
    - name: F1
      type: f1
      value: 0.774
---

# Rumour Detection
You can **test the model** at [Rumour-Detection-Twitter](https://huggingface.co/spaces/aisingapore/rumour-detection-twitter) | [SGNLP-Demo](https://sgnlp.aisingapore.net/rumour-detection-twitter).<br />
If you want to find out more information, please contact us at sg-nlp@aisingapore.org.

## Table of Contents
- [Model Details](#model-details)
- [How to Get Started With the Model](#how-to-get-started-with-the-model)
- [Training](#training)
- [Model Parameters](#parameters)
- [Other Information](#other-information)

## Model Details
**Model Name:** Rumour-Detection
- **Description:** This model is based on the hierarchical transformer architecture described in the associated paper.
- **Paper:** Interpretable rumor detection in microblogs by attending to user interactions. Proceedings of the AAAI Conference on Artificial Intelligence, April 2020 (Vol. 34, No. 05, pp. 8783-8790).
- **Author(s):** Khoo, L. M. S., Chieu, H. L., Qian, Z., & Jiang, J. (2020).
- **URL:** https://ojs.aaai.org//index.php/AAAI/article/view/6405

# How to Get Started With the Model

## Install Python package
SGnlp is an initiative by AI Singapore's NLP Hub. They aim to bridge the gap between research and industry, promote translational research, and encourage adoption of NLP techniques in the industry. <br><br> Various NLP models, other than aspect sentiment analysis are available in the python package. You can try them out at [SGNLP-Demo](https://sgnlp.aisingapore.net/) | [SGNLP-Github](https://github.com/aisingapore/sgnlp).

```python
pip install sgnlp

```

## Examples
For more full code (such as Rumour Detection), please refer to this [SGNLP-Github](https://github.com/aisingapore/sgnlp). <br> Alternatively, you can also try out the [Rumour-Detection-Twitter](https://huggingface.co/spaces/aisingapore/rumour-detection-twitter) | [SGNLP-Demo](https://sgnlp.aisingapore.net/rumour-detection-twitter) for Rumour-Detection-Twitter.

# Training
The train and evaluation datasets were derived from the Twitter15, Twitter16 and PHEME datasets. The full dataset can be downloaded from the author's [Dropbox](https://www.dropbox.com/sh/w3bh1crt6estijo/AAD9p5m5DceM0z63JOzFV7fxa?dl=0).
- **Training Config:** Not available

#### Training Results
- **Training Time:** ~6 hours on a single V100 GPU.

# Model Parameters
- **Model Weights:** [link](https://storage.googleapis.com/sgnlp-models/models/rumour_detection_twitter/pytorch_model.bin)
- **Model Config:** [link](https://storage.googleapis.com/sgnlp-models/models/rumour_detection_twitter/config.json)
- **Model Inputs:** Thread of tweets. The first tweet should be the target tweet.
- **Model Outputs:** Array of logits for each class (True, False, Unverified, Non-Rumour). This can be converted into probabilities using the softmax function.
- **Model Size:** ~60mb
- **Model Inference Info:**  Not available.
- **Usage Scenarios:** Rumour detection / fake news detection on Twitter

# Other Information
- **Original Code:** [link](https://github.com/serenaklm/rumor_detection)