Document classification utils

This project contains some useful tools for various document classification tasks. More specifically ...

Running from Docker

To run the project as a Docker container you first need to prepare the required data. In particular, prepare a folder that contains

• models folder with the model files
• label_mappings folder including the JSON files (one per language, it.json for Italian) containing the descriptions of the EuroVoc labels
• id2label.json JSON file containing the descriptions of the IPZS
• i2eu_id.json JSON file containing the mappings between IPZS and EuroVoc codes
• mt_labels.json JSON file containing the MT and DO parents for EuroVoc

To execute the docker container, run the following command:

docker run -p 80:80 -v /path/to/data/folder:/data/ smartcommunitylab/document-classification

where /path/to/data/folder should refer to the folder with the above data.

Once started (may require quite some time due to the size of image and the models), it is possible to

• use Web interface available on http://localhost/ui
• consult the API documentation of the server http://localhost/redoc

Creating Docker image

To create a new version of the Docker image, run the following command from this directory

docker build -t imagename .

replacing imagename with the name of your choice.