Milestone 3 ABC DarkHorseIdea: Trust based recommenders

From crowdresearch
Jump to: navigation, search

Using Trust Based Recommenders for recommending Tasks

The problem with most crowdsourcing platforms is that the jobs which are available to the new workers are not according to their preferences. Also, because of this, the workers have to waste a lot of time in searching a relevant job for them. So, in order to address this problem, we could think of a mechanism which will allow the workers to have access to jobs which they have interest in and also, at the same time, they trust the requester who posted the job. This we believe could be done by building a Trust Based Recommender System. The system is going to have the following specifications:

1. In order to create trust among workers for the requesters, the data from social network sites could be used so that the tasks which will be recommended to the Workers will be based on the rating given by their peers/family on Social Media. This has been thought because people usually tend to trust their friends and relatives more than anyone else. We could create a network of people who the workers are most close to and use the inputs from these people for task recommendation.

2. Also, apart from trust, another parameter that could be used is how much skills and interest does a person share with the Worker. So, our aim is to recommend jobs to users which have been highly rated by someone who has a lot of commonalities with the worker. This way the skills which the user has will be required in the tasks allowing him to improve his skill set and also his interest will be maintained, thus preventing the job from becoming laborious.

3. Another advantage of Trust Based Recommender System is that it works well on sparse data which we believe will be the case in recommending jobs on a platform like MTurk.