WikiRecommender
Wikipedia Page Recommending System
Copyright (C) 2016 Szymon Lis email: [email protected] twitter: @drSzymonLis
This program is distibuteted for free but WITHOUT ANY WARRANTY
Description
Wikipedia Recommender is a Python script using a machine learning techniques for retrieval the most similar wiki page to query article (query does not need to be a wiki page) The project was started in December 2016 by Szymon Lis. Project is still in early stage of development.
Installation
Dependencies
- Python (>= 2.7) [It might aslo work with Ptython >= 3.3]
- NumPy (>= 1.6.1)
- Pandas (0.18.1)
- Kitchen (1.2.4)
- ElementTree (1.0.1)
User installation
Script needs a wikipedia dump file in xml format. You can download it from: https://en.wikipedia.org/wiki/Wikipedia:Database_download Install all dependenices berfore starting the script.
Communication
Autor: Szymon Lis [email protected]
Realse
24/01/2017
- k-nearest neighbor algorithm
- euclidean distance metric
- word count
Issues
For big data transformation into sparse matrix takes huge amount of time and I'm currently working on speeding up this process. There is possibility to store precumpoute data on disk to avoid heavy computation. Also k-nearest neighbor works really.