tanaos-spam-detection-italian: A small but performant base spam detection model specifically designed for Italian text

This model was created by Tanaos with the Artifex Python library.

This is a spam detection model based on distilbert-base-multilingual-cased and fine-tuned on a synthetic dataset to classify text as spam or not_spam.

This model is intended to be used as a first-layer spam filter for email systems, messaging applications or any other text-based communication platform, and it was specifically fine-tuned to perform well on Italian text.

The following categories are considered spam:

1. Unsolicited commercial advertisement or non-commercial proselytizing.
2. Fraudulent schemes. including get-rich-quick and pyramid schemes.
3. Phishing attempts. unrealistic offers or announcements.
4. Content with deceptive or misleading information.
5. Malware or harmful links.
6. Adult content or explicit material.
7. Excessive use of capitalization or punctuation to grab attention.

Languages

The main model language is Italian, but we have spam detection models specialized in other languages as well:

• English: tanaos-spam-detection-v1
• Spanish: tanaos-spam-detection-spanish
• German: tanaos-spam-detection-german

How to Use

Use this model through the Artifex library:

install Artifex with

pip install artifex

use the model with

from artifex import Artifex

spam_detection = Artifex().spam_detection(language="italian")

print(spam_detection("Hai vinto un iPhone 16! Clicca qui per ottenere il tuo premio."))

# >>> [{'label': 'spam', 'score': 0.9989}]

Model Description

• Base model: distilbert/distilbert-base-multilingual-cased
• Task: Text classification (spam detection)
• Languages: Italian; for other languages, see:
  • English: tanaos-spam-detection-v1
  • Spanish: tanaos-spam-detection-spanish
  • German: tanaos-spam-detection-german
• Fine-tuning data: A synthetic, custom dataset of spam and not spam examples.

Training Details

This model was trained using the Artifex Python library

pip install artifex

by providing the following instructions and generating 10,000 synthetic training samples:

from artifex import Artifex

spam_detection = Artifex().spam_detection()

spam_detection.train(
    spam_content=[
        "Pubblicità commerciale non richiesta o proselitismo non commerciale",
        "Schemi fraudolenti, inclusi schemi per arricchirsi rapidamente e schemi piramidali",
        "Tentativi di phishing, offerte o annunci irrealistici",
        "Contenuti con informazioni ingannevoli o fuorvianti",
        "Malware o link dannosi",
        "Contenuti per adulti o materiale esplicito",
        "Uso eccessivo di maiuscole o punteggiatura per attirare l'attenzione",
    ],
    language="italian",
    num_samples=18000
)

Intended Uses

This model is intended to:

• Serve as a first-layer spam filter for email systems, messaging applications, or any other text-based communication platform, if the text is in Italian.
• Help reduce unwanted or harmful messages by classifying text as spam or not spam.

Not intended for:

• Use in high-stakes scenarios where misclassification could lead to significant consequences without further human review.