An IA model that detects whether a given verse is from the Bible or not
The model presents capabilities at a very high recognition level, for the Hebrew language.
The complete dataset on which the model was trained is stored in the bible_data.csv
file.
You can try the model's capabilities easily,By downloading the release file from here - https://github.com/NHLOCAL/is-this-bible/releases/download/v1.0/is-this-bible.zip.
Want it completely easy? Try the model in the example space - https://huggingface.co/spaces/NHLOCAL/is-this-bible
To run the model, download the following libraries using pip:
nltk
, joblib
.
example:
Negative input:
try_model.py "בגיטהאב ניתן להעלות מערכות קוד פתוח"
output:
Text: בגיטהאב ניתן להעלות מערכות קוד פתוח | Prediction: Other | Confidence Score: 0.0340
Positive input:
try_model.py "עניה סערה לא נחמה הנה אנכי מרביץ בפוך אבניך"
output:
Text: עניה סערה לא נחמה הנה אנכי מרביץ בפוך אבניך ויסדתיך בספירים | Prediction: Bible | Confidence Score: 1.0000