WinterMilestone 5 BITS

From crowdresearch
Jump to: navigation, search

System For Task Feed

Creating a well matched and efficient task feed is a complex task and depends on many factors , a major part of our idea last week -> (Ranking the tasks by categorizing them on the basis of tags provided by the requesters) is the correct analysis of the skills of a worker , the boomerang works best for providing trustworthy reputation scores which can also be extended for other attributes , boomerang can also be used to make the requester tag a worker on a certain set of attributes . Finding a good way for permuting the tasks is essential for the crowdsource-platform.


We introduce a new concept much like boomerang , the main aim of our System is to allow for efficient and accurate tagging of workers for particular attributes , our System uses these features:

1)Endorsement capabilities wherein a worker/requester can endorse a worker for a particular skill.

2)Requesters face the wrath of boomerang if they endorse the workers wrongly.

3)For the better endorsement of workers by other workers , we have conceptualized an endorsement score , which would differentiate good endorser and bad endorser.A worker endorsed by good endorser for a skill will be more likely efficient in the skill than the worker endorsed by a bad endorser . The endorsed skills would be bound to the present endorsement score of the worker i.e

A-> worker , 
B->worker , 
A endorsed B for skill X ,
A has M(good) reputation score in 2015 ,
A's reputation score dropped to N(bad) in 2016 ,
So the worker B's skill X would be rated on the basis of A's present reputation score and not the previous one(which was good).

4)A good endorsement , agreed upon by the requester would result in a non linear growth in the endorsement scores of the worker who endorsed the worker.

How is the system solving critical problems

The system solves a major problem which is non trustworthy tagging of workers for skills by other workers, this system can be adopted for promotion of open governance idea as well , it emphasizes on the involvement of worker community in determining your skills in turn of the risk of getting their endorsement scores dropped . Thus a worker would only endorse other worker when that worker is truely skilled. Once the workers are correctly tagged , and the requesters have correctly tagged the tasks , it only takes the need of a matching algorithm which would then form clusters , where each node would contain two fields , first field -> requester/worker, second field-> a set of skills , reputation scores , attributes.

Now if we look at this system , this is a bipartite matching problem in a sense , as we have two categories of nodes and we intend to find a perfect matching between them !


The System aims to create a good pairing between workers and requesters, and also provides a good way for creating a well permuted task feed.

Module 1:Tagging functionality

Creating an interface and functionality for endorsing workers based on their skills,reputation etc. where skills and attributes will be tagged using a predefined set of skills and attributes database . Endorsing a worker for a skill should send a notification to the worker who is endorsed .

Module 2:Endorsement Score

The system aims to provide correct endorsements for workers , this would result in differentiation between good quality workers and bad quality workers posing as being skilled. Workers can be endorsed by requesters and Workers both , Requesters would tag workers based on their skills, work quality and as a boomerang effect , their task feed would be available to that particular worker earlier .Whereas if workers tag other workers , that will be done using an interface which allows workers to tag other workers on skills and work quality. A worker can endorse another worker on multiple skills , but has to risk his/her endorsement scores which would define how good a worker is in the community. Endorsement scores and reputation scores combined can provide a good ranking and judgment about an individual worker. So good endorsement scores should also affect the worker i.e a endorsement score can be used to directly affect the reputation score for example.

Module 3:Snippets of good work quality

This module aims to help inefficient and confused workers which do not understand or satisfy the requester's expectations , to get good quality of work , requesters can provide snippets of good quality work done by other workers on the requester's task , adding a workers work in requester's profile would result in similar boomerang effect i.e the requester's task would be more available to the worker whose work , he has added in his good quality snippets. This little feature provides workers with a rough but close idea on how to perform the task and create a good piece of work in accordance with the good quality snippets provided by the requester. This feature aims to train and supervise inefficient workers and raise their skill levels.

Module 4:Filters in task feed based on endorsed skills

The tasks in the feed of a worker must be filtered based on the skills for which he/she is endorsed, this provides a good and sure proof method for getting those tasks which would surely be good for the worker as he is skilled in those tasks.


@shivam and @shivangi github-> Curious72(shivam) and insomniac12(shivangi)