Recommender Engine for WikiMedia proof-of-concept system.
It uses Ruby lang and Mongo DB (it was 2013 😬)
It consists of:
- batch jobs that prepare data index to be quickly looked up.
- API server that uses the index and serves responses to JS frontend. - JS frontend standalone from wikipedia - for semantically exploring WikiMedia content
- a JS plugin for WikiMedia PHP for looking up similar articles withim WikiMedia
The system was meant to handle security related topics so there is some security specifc code out there, especially on the frontend.
For configuration there is a list of stop words that should be ommited when building search index.
The list is in ./lib/text-mining.rb.
Polish language has also bunch of sufixes and conjugation that the text-mining.rb can take care of.
But it's very language specifc.
There is a Mongo DB index connector and WikiMedia MySQL connector inside ./lib folder.
jobs directory has all the code for batch processing jobs.
Rakefile - has description of all the possible tasks
The API server that uses the index prepared by batch processing jobs is inside ./apis directory.
Has a lot of poorly written polish language ridden code.
The frontend code is placed somewhere between ./frontend and ./apis/public
The WikiMedia JS plugin is present in ./apis/public/wiki-plugin.js - ish
It's in the LICENSE file