Link Prediction for PIR

Private repo of the DD2477 - Search Engines and Information Retrieval Systems project at KTH.

Done by Titouan Mazier, Lorenzo Sibille, Tristan Perrot & Hani Anouar Bourrous

Data

The data used in this project can be found and downloaded here.

Connect to ElasticSearch

Create the file connector.py, which should follow this template

from elasticsearch import Elasticsearch

def establish_connection():
    return Elasticsearch('https://localhost:9200',
                           basic_auth=("elastic","YOUR PASSWORD HERE"),
                           ssl_assert_fingerprint="YOUR SSL KEY HERE")

Use the values you get when performing your installation. Multiple connetcion options are possible, see ElasticSearch documentation here.

Python packages

Before running the code, be sure to have the required Python Packages. The complete list is in requirements.txt.

Running

To run the code run from the terminal you need to launch an docker instance of Elastic Search and run PIR.py file.
Flags:

• -d \[AOL\] to change the dataset. If not specified, AOL4PS is used.
• -l to index the chosen dataset into ElasticSearch. Should be done the first run only.
• -n <integer> to specify the maximum number of files you are interested in.
• -v to run the code in evaluation mode. It will compute and print metrics about the reranking process, without running the GUI.