Wikibooks:Reading room/Proposals/2019/February

A New Recommender System For Wikibooks (can be extended to WikiMedia in general)
I created a phabricator task for the proposal to improve the existing recommender system used by WikiMedia as a project for GSoC 19 and a mentor asked me to share my proposal here. My proposal includes the followinng tasks:
 * Making a form at the end of each article, so that user can tell if he liked the article or not (maybe star based system) and the categories he wants to tag the article with.
 * Based on this star based system, and the tags provided by users, it will learn new categories for the article.
 * It will recommend new articles, after some time of learning, to users based on their previous activity, in parallel to present system based on edits made by users.
 * learning similarity between different users based on their interests and then it will try to suggest discussion pages where they can discuss similar ideas.

following is the idea on which the proposal is based: “Some people use Wikimedia to gain knowledge rather than making edits and improving what already exists.”

My idea of recommender system serves people who surf Wikimedia to learn new things along with those who know a particular topic and mostly visit articles surrounding that topic only and it will help people to discover new people who are interested in similar topics and in this way we can become part of larger group dynamics and discussions.


 * Current recommender system:
 * The current recommender system is mostly based on the edits made by the user. So, this method doesn’t serve those who want to learn new things and have no prior knowledge about the articles they want to visit.
 * The second method on which current recommender system works is by user providing their interests.
 * In the current recommender system, there is no way by which the system can learn new categories that should be assigned to the page and which may be skipped out by the person who has tagged the article with categories.
 * New idea of recommender system:
 * This recommender system learns categories liked by new users. The recommender system suggests new articles to users based on these categories assigned to new users.
 * It will also learn categories to be assigned to new articles based on the ratings given by old user.
 * It will adapt to changes in interests of users. If a user is interested in new type of categories, it will change the categories assigned to user automatically.
 * We can also add a feature that helps a user assign new categories to articles. This will help us to track user’s interests and change categories liked by him accordingly and it will help learning the categories of articles easily. This will also help us to tackle the emergence of new categories.
 * By knowing similarities between users, we can make them part of larger discussions.
 * Also, we can ask new users about their fields of interest and show initial results (results before learning) accordingly.
 * This recommender system can run in parallel to the old one. So, some of the suggestions will be based on this system and others will be given by old one.

(User:Chaitanya kharyaldiscuss) 18:09, 17 February 2019 (UTC)