Summary by Ioana

CrowdDB: answering queries with crowd sourcing.     - Michael J. Franklin,
Donald Kossmann, Tim Kraska, Sukriti Ramesh, Reynold Xin

The paper presents a relational query system ( CrowdDB ) that uses
crowdsourcing in order to answer queries that otherwise cannot be answered ,
due to incomplete data or subjective comparison. The system extends SQL DDL
and DML with a number of operators, in witch human input is considered and is
obtained by generating small task for crowdsourcing platforms.

The experiments presented in the paper are ran in October 2010, over 25.000
HIT on AMT and measure the response time and quality of the answer. The main
findings are : - the HIT Group size  influences the performance of the group (
the response time of the same query decreases when the size of the HIT Groups
increases); - there is a tradeoff between throughout( HITS competes per unit
of time) and completions percentage; - ( an obvious ) increase in the payment
of the task increases the quality of an answer ( in the example from 1 cent to
4 cents ) ;


Questions : 1. How would a quality feature be integrated in CrowdDB?( quality
of response )  2. How reliable is the response obtained? Does the availability
of the operators influence the response ?


Summary by Patrick

CrowdDB is an attempt to answer queries using crowdsourcing. The idea is to
use crowdsourcing to handle problems that are too expensive to either solve or
express for a machine to solve.

Amazon Mechanical Turk is an online system allowing humans to solve HIT (Human
Intelligence Tasks) for a pre-defined payment.

The authors propose a DBMS where queries can be converted to a user interface
which turks can fill out missing data. The experiments seems fairly decent.




Summary by Zahid

CrowdDB: Answering Queries with Crowdsourcing

Michael J. Franklin, Donald Kossmann, Tim Kraska,
Sukriti Ramesh, Reynold Xin

Summary

All quires are not answerable through machine so some of them need human
interaction or feedback to get answers. These queries have not answers in
databases. So to get answers of these queries authors worked on CrowdDB that
uses human input through crowdsourcing. The traditional  closed-world
assumption for query processing does not hold for human input. For a human
oriented queries operators are needed to integrate the crowdsourced data. For
a high performance and good results it need workers training , worker
affinity, motivation, and locations, all these factors can get through heavy
cost. Traditional data bases try to answer the query according to information
that it has, and if there is not information regarding to any query than it
consider no answer. So that¡¯s why human interaction or intervenes need to
understand the query and try to answered it.  The main contributions of this
paper is as follow.

They proposed simple SQL schema and query extensions that enable the
integration of crowdsourced data and processing.

They present the new design of CrowdDB having new crowdsourced query operators
and plan generation techniques those are helpful to combine the crowdsourced
and traditional query operators.

Authors described different methods to generate the effective and
automatically user interfaces for crowdsourced tasks.