Abstract – The Government of India and various state governments have been offering several poverty alleviation program and schemes to benefit BPL (Below Poverty Line) families. Though the intention of these schemes is very good, implementation and management of these are not up to mark. It has been observed that the main reason of their ineffectiveness is manual intervention that leads to either exploitation of a BPL person or funds going to the wrong

persons. The BPL card system proposed by us is targeted to reduce this problem as much as possible. Under this Card system, each intended beneficiary will be issued an identification Card called “BPL Card” containing machine-readable information about the beneficiary including eligibility under various schemes. Special Kiosks manned by trained operators will be provided within easy access of the Tehsil/Taluka of each beneficiary. Using the BPL card the beneficiary will be able to obtain information about the various schemes as well his eligibility to apply for any or some of them. 

1. PROBLEM STATEMENT

Background: The  different State government operates several poverty alleviation (anti-poverty) schemes and projects for the up liftmen of economically backward regions and citizens of that state. These schemes are funded by various agencies including the Central Government of India as well as the State Government. Each scheme is intended to address specific

needs of defined sections of the population – Housing Finance Aid schemes, Employment Generation schemes, Infrastructure development schemes, Health promotion schemes, as well as special schemes for Scheduled Castes and Tribes and members of the Fishing communities.

The effectiveness of the implementation of these schemes is considerably limited by various factors. The Department of Social Justice & Empowerment has conceived of a special

system to ensure that the benefits of the various schemes reach the intended recipients. An additional benefit of the system will be the easy availability of small amounts of state-funded,

unsecured low-cost credit to the individuals covered by the scheme.

Introduction: Under this system, each intended beneficiary family will be issued card called “Below Poverty Line Card” (BPL Card) containing machine-readable information. This card is low cost card and having unidirectional communication. We use a RFID Card reader. This Card

reader red the unique 16bit/32bit numeric data embedded into it and transmit to server. Server searches the database and provides the information. It will contain the name, age, category and address of the Head of the family and the validity period of the card, along with detail of his/her family members including eligibility under various schemes. It is assumed that a BPL person may not be literate enough or skilled to operate this card at any kiosk. So it is thought that the kiosk be managed by a trained operator. Special Kiosks manned by trained operators, will be provided for the help of beneficiary.

By using this BPL card the beneficiary will be able to obtain information about the various schemes provided by different government and non government organization for him and

also apply for them. The Goal of the BPL Card Scheme:

· No any need of filling up of any type of application forms.

· For the benefit, card holders have no need to any intervention of the Gram Sabha, Gram Sevak, Tehsil Officer and the other Government Machinery. So, they got more time for other  development work.

· Provide Identification of the BPL population 

· Control of miss-utilization of benefits

The objective of the BPL Card Scheme:

· Easily Operable System Access to a fixed amount of credit (as a part of different government schemes like “Rashtriya bridha Pension”, “Rajya Samajik Suraksha Pension”, “Widow Pension”)

· Access up to some fixed amount of money at a fixed interest rate as loan, at any time.

· Mobility (Access to the s ystem across that state)

· Least processing time

This BPL card will be used by different Government and non Government organization for schemes covering:  

· Housing Finance Aid schemes

· Employment Generation schemes

· Health promotion schemes

· Small Value Individual Schemes

· Different type of Pension schemes

· Public distribution system (In future Government may combine PDS Card with BPL Card)

Scope : RFID-enabled cards are becoming used in many other contexts such as public transportation system [1]. The New York City subway recently in 2006 started a trial of RFIDenabled cards for transportation [2]. A participant in this trial uses his/her credit card as a transit ticket in place of the traditional magnetic stripe-based dedicated subway ticket.

BPL card may also be thought to be used in various such ways so that the community belonging to BPL category (possibly content, provide the end user with a memorable

experience.

 Centralized data collection: There is a strong need for a common repository for the data obtained from the collaborative applications. Also centralized collection and management of data across applications provides a background for analysis that can help identify better designs, and isolate design patterns.

 Data collection from multiple sources: There is a lot to learn from the collective behavior of users over the internet. Much richer information can also be obtained from the interactions between people if there is a reliable and consistent method of collating data from multiple

sources. 

 Scalability: The applications must be capable of running on an internet scale.

The applications would be built upon the Microtasks framework – which is based on the technologies being developed by the Advanced Development and Prototyping Group at MSR India.

Building a Microtasks application falls within the domain of Web 2.0 and would involve both experimentation and prototyping in C#, Silverlight, ASP.NET AJAX, HTML, Javascript and XML Web services.

RELATED WORK

I. Human-Based Computation Traditional computation involves a human programmer who

describes the problem formally to a computer and receives a solution based on his logic and the computers calculation capability. In the domain of Human-Based Computation, we find that it is the Computer who asks the human to solve a computation problem and subsequently consumes the solution in a constructive manner.

Essentially, human-based computation is a technique when a computational process performs its function via outsourcing certain steps to humans [4]. This approach exploits the fact that there are still some computational problems that can be solved by humans and not by computers.

In the research conducted by Caldwell et al [5] and Sims [6], humans would utilize their visual perception and aesthetic judgments to train an evolutionary computer algorithm. After

sufficient training, the computer program was capable of evolving beautiful faces and brilliant works of art.

II. Games with a Purpose In his work titled „Games with a Purpose‟ [7] von Ahn suggests that each year, people around the world spend billions of hours playing computer games. What if all this time and energy could be channeled into useful work? What if people playing computer games could, without consciously doing so, simultaneously solve large-scale problems? This

again relies on the fact that there exist a subset of problems which a computer cannot solve, but it might be trivial for a human. But one major difference lies in the fact that unlike computer processors, humans require some incentive to become part of this collaborative computation.

Online games are an effective way for encouraging people to participate in the process. Such games constitute a general mechanism for using brain power to solve open problems.

These games with a purpose have a vast range of applications in areas as diverse as security, computer vision, Internet accessibility, adult content filtering, and Internet search. A few

such games under development as described below can solve problems that computers can‟t yet solve. 

 ESP Game: A game in which two randomly paired people are simultaneously given the same image, with no way to communicate among one another. Each then lists a number of words/phrases that they feel describes the picture within a time limit. If and when the users agree on a label, we can be sure that it is an accurate description of the picture, as both parties acknowledged it. Currently Google Inc. has licensed the ESP game in the form of the

„Google Image Labeler‟ wherein users, while fighting for points and top honors, are labeling the images of the World Wide Web.

 CAPTCHA: The Completely Automated Public Turing test to tell Computers and Humans Apart (CAPTCHA) can be used to identify whether one is interacting with a human or a computer. The user is presented with a warped image containing text, unreadable by modern

character recognition algorithms, and asked to write the text within the image. The More recently, von Ahn developed reCAPTCHA, a novel approach to CAPTCHA wherein instead of a generated image sequence, the reCAPTCHA consists of text that could not be recognized

by an OCR mechanism, during the digitization of books. 

On solving the reCAPTCHA, the user essentially is proofreading the digital books.

 Peekaboom: Peekaboom is game for locating objects within images. Player one is given an image and a word (the pairing is output from the ESP game); the player clicks on the image, and reveals a small portion to be shown to the other player, who then has to guess the word, based on what he sees. On correct identification, we obtain the accurate position of an object in an image.

Once this game generates a large enough data set, it could potentially be used for training computer vision systems. 

 Verbosity: This game collects common-sense facts ("Milk is white", "Milk goes with cereal.") A player is given a word, and fills in a phrase to describe it. The other player is given the word, and has to guess the same descriptive phrase. This is known as an Asymmetric game as the players play different roles. Meanwhile, the ESP game is symmetric as both players perform the same task.

There is a lot of potential in developing such games oriented towards machine learning research. If we can make the games fun, we are guaranteed to attain a wide audience of casual gamers over the internet, hence obtaining large data sets.

3

III. Amazon Mechanical Turk Developed and hosted by Amazon‟s huge grid infrastructure,

Amazon Mechanical Turk [4] provides a set of infrastructure to post and farm out what they term as “Human Intelligence Tasks”. It differs from the Microtasks framework that we used, in one aspect. Mechanical Turk is not generic enough for generating interacting applications where multiple humans can collaboratively generate some useful output.

TECHNOLOGY USED

I. Microtasks Framework Our aim is to abstract the interaction of the Application and

the User so that it passes through the Micro-Tasks infrastructure. One hopes that by reducing these direct interactions; one will make the task of the application developer easier. Essentially the user performs the jobs, and the application developer provides specifications for the tasks

and also the user interface. The MT infrastructure sits in between and actuates the Application based on the actions performed by the users, and also provides the correct UI to the users as specified by the application. It also mediates the interaction of the user and the application, including the task assignment.

II. Silverlight

Microsoft Silverlight [9] is a web-based, cross-browser, crossplatform presentation foundation. Silverlight (code-named Windows Presentation Foundation/Everywhere or WPF/E) is a

proprietary runtime that allows Rich Internet Applications (RIA‟s) to run within the confines of web-browser. Some of the features offered are: user interface development, 2D drawing, fixed and adaptive documents, advanced typography, vector graphics, raster graphics, animation, audio and video. Silverlight is based on eXtensible Application Markup Language (XAML) and JavaScript. The Silverlight subset enables Flash-like web and mobile applications with the exact

same code as Windows .NET applications. 

CURRENT WORK

Reading text from images has always been a challenge for computer vision researchers. Not only are the recognition algorithms extremely complex but also there is little verifiable data to confirm the working of the system and to train it for future executions.

I have been working on an Image Annotation Application, which proposes a solution to the abovementioned problem. A researcher can define a task as a set of images that he wishes

to obtain annotations for. He can then use this application and the underlying Microtasks framework to widely distribute this task to several casual users or volunteers.

The users, or in this case the „actors‟ related to this task are required to mark-out parts of the image which contain text and label, using the keyboard input, what the text within the

image represents. If this task is repeated over a large data set, we can obtain accurate information about the text within the image and train self-learning computer vision algorithms to recognize the text, the next time it encounters it.

ORIGINAL IMAGE AND IMAGE ANNOTATION APPLICATION

The current version of Image Annotation has been implemented using Microsoft Silverlight (WPF/E) for the frontend and UI. The XML web services are implemented in ASP.NET and the server side code uses C# on the .NET 3.5 framework.

CONCLUSION & FUTURE WORK In the 50 years since term artificial intelligence became

commonplace we have seen gargantuan leaps in the ability of computers to mimic human intelligence. Given these advances, the computer still fails to complete some tasks

which humans can perform intuitively.

Capturing this innate ability, within the notion of Human  Computation, has the potential to not only analyze and correctly verify vast amounts of data but also help evolve machines to become more intelligent. For the remainder of my B.Tech. Project I hope to explore various other instances of Human Computation and large scale collaboration. I would, side-by-side, be designing and developing applications, based upon the Microtasks framework, which would serve as corpora collection tools for the researchers within Microsoft. We are in the very early

stages of building our next application, a collaborative audio/video annotation tool, which would allow multiple users to chat and interact with one another, while annotating an audio/video clip with text (such as lyrics, subtitles or transcripts). We aim to deploy this application in partnership with the Speech Recognition research group, in order to collect large amounts of training data.