Project contain methods and classes that can help to create classification model (based on Keras+TensorFlow) for URLs/pages/documents that can be with different languages.
Also, project have pre-trained model for filtering URL/web-pages with adult content. This model receiving title, description, keywords and body text as inputs and return "Normal" or "Adult" based of received text. Model can classify cs, de, en, es, fr, hu, it, nl, pl, pt, ru, tr, uk languages.
Accuracy of a model - 95-99% (depending on language).
Prediction time - ~0.0045 sec per document.
Pre-trained model can be use via console and like a module in another project.
Console command:
python --title <string> --keywords <string> <string> ... <string> --description <string> --body-text <string>
where:
--title - title of a web-page;
--description - description of a web-page;
--keywords - keywords of a web-page;
--body-text - body text of a web-page.
For usages in other modules, please, look at example.py.