Scientific papers recommendation system

This project aims to recommend relevant scientifi papers to a researcher, according to her field of expertise and her interests.

To do that, we use deep learning techniques like Skip-Thought vectors and DSSM.

The documentation of this project is available at: https://tizot.github.io/recom-system

Installation

The project is developed in python 3. We used specifically the version 3.4.

Requirements

We recommend to use a virtualenv to keep projects separated on your machine. Obviously, this is not mandatory.
To install the project, firstly clone the repository from Github, then install python dependencies.

git clone https://github.com/tizot/recommendation-system.git recom
cd recom
virtualenv --python=python3.4 .env
source .env/bin/activate
pip install -r requirements.txt

If you do not want to use a virtualenv, install the following python packages:

• numpy 1.11.0
• scipy 0.17.1
• pandas 0.18.1
• matplotlib 1.5.1
• mysqlclient 1.3.7
• scikit-learn 0.17.1
• Theano: pip install --user https://github.com/Theano/Theano/archive/master.zip
• Lasagne: pip install --user https://github.com/Lasagne/Lasagne/archive/master.zip

SQL database

In order to use the scripts, you need a SQL database in which are stored all the papers. You can name it as you like, the default name in the script is dblp.

In this database, you must have a table called papers with at least three columns:

• id: a unique identifier for each paper (INT or UUID);
• title: the title of the paper (VARCHAR(255));
• abstract: the abstract of the paper, that can be empty (TEXT).