This project, developed by the DARElab team at the ATHENA Research Center, is part of the FAIRCORE4EOSC European project. Its aim is to support research discovery within the RDGraph by providing different recommendation services to users.
This repository contains the source code under darelabdb/
and the wheel file (fc4eosc_recommenders-0.1.0-py3-none-any.whl
) for the recommender systems integrated into the RDGraph portal.
git clone [email protected]:athenarc/fc4eosc-recommender.git
cd fc4eosc-recommender
pip install fc4eosc_recommenders-0.1.0-py3-none-any.whl
After installation, create a .env
file in the root directory by following the structure in .env.example
.
After importing a new database dump, a few setup steps are required:
- Add sk_ids as primary keys instead of the string ids in the database for better performance (
./sql/add_sk_ids.sql
) - Add trigram-based indexes to string fields (
./sql/build_extra_indexes.sql
) - Create the recommender schema (
./sql/build_recommenders_schema.sql
)
You can automate these steps using the provided initialization script:
cd sql;
chmod +x initialization.sh;
./initialization.sh <host> <port> <user> <password> <database>
There are 3 different recommenders available:
User-to-Item Recommender Recommends research products (e.g., publications) to users based on their ORCID and their citation behavior within research communities.
Item-to-Item Recommender Recommends similar research products based on metadata and textual similarity.
Category-based Recommender Recommends research products based on category associations.