The tool is available at http://figureout.cs.technion.ac.il/.
Use it to detect metaphors in Hebrew poetry.
The metaphor detection is based on a machine learning model that was trained on a dataset of Hebrew Piyyut poetry. Details about the model and the dataset can be found in the paper A Dataset for Metaphor Detection in Early Medieval Hebrew Poetry.
- Metaphor detection based on textual input (Hebrew).
- Support for multiple texts in a single request.
- Metaphor detection based on pre-defined Hebrew Piyyut corpus.
- Side by side comparison of model annotations and user corrections.
- Showing key statistics about the model's performance (precision, recall).
- Download annotations in JSON format.
- Dynamically adjust annotations based on a threshold.
- The backned is based on Flask.
- The frontend is based on Vue.js.
- Remember to create .env file in each environment (linux, windows).
- vue.js setup is explained in this tutorial.
- Execute the Flask server:
authbind --deep python app.py
- Execute the Vue.js server:
authbind --deep http-server -p 80 dist/