---
license: cc-by-nc-4.0
datasets:
- ucirvine/sms_spam
- ealvaradob/phishing-dataset
- Deysi/spam-detection-dataset
- webis/tldr-17
- KaiNylund/twitter-year-splits
language:
- en
base_model:
- distilbert/distilroberta-base
pipeline_tag: text-classification
library_name: transformers
tags:
- pytorch
---
# Investi-gator's V2.0 Scam/Spam Classifier (distilRoBERTa)

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

On social media, it's very easy to run into posts that may not be authentic. This makes some question if the content their consuming is real or legitimate. For others it can lead them to consume misinformation or engage with users and content that don't have their best interest at heart. That's when Investi-gator steps in to help users take back control of their social media experience, using this model to classify scam/spam content.


### Model Description

<!-- Provide a longer summary of what this model is. -->
This model is specifically meant to classify scam/spam patterns in a general or social media setting. 


- **Developed by:** [bthomas279]
- **Model type:** [More Information Needed]
- **Language(s) (NLP):** [English]
- **License:** [cc-by-nc-4.0]

### Training Scope 
Here's a list of the types of data used to train V2 of the Scam/Spam Detector:
- Phishing
- SMS
- Real V1 False positives (on Reddit and Twitter. Varying topics)
- Engagement farming
- Crypto pump
- Text with clickbait links
- Fake livestreams
- Sponsered-ad-style scam
- Affiliate/dropshipping spam
- Fake discounts
- Impersonation
- Fake givaway or charity
- Fake customer support


## Bias, Risks, and Limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->
Known Biases (Positive leaning(scam/spam):
- Promotions in general
- Texts mentioning money
- Advertising