Product labelling

Application for labelling transaction receipts, developed as part of the transaction data project. Predictions are made using the fastText algorithm.

The application available at ~/home includes two parts :

• The main part aims at simplifying the labelling process. One label from a file are automatically offered to the user who has to choose the rigth associated category/label.
• The second one enables users to visualize the results (preprocessing and predictions of the model).

Quick start

Using docker

docker run --env inseefrlab/product-labelling

Configuration

Each variable can be overriden using environment variables.

Product-labelling configuration

Key	Default	Description
model	none	URL of text classification model - fasttext model saved with .ftz extension (must be configured)
nomenclature	none	URL of a CSV file which contains complete list of nomenclature products with no header
db_type	sqlite3	Other supported mode : postgres

Product-labelling configuration if dbtype==postgres

Key	Default	Description
db_password	none	See django configuration (must be configured)
db_name	none	See django configuration (must be configured)
db_user	none	See django configuration (must be configured)
db_host	localhost	See django configuration
db_port	5432	See django configuration

Documentation

See Django documentation