a market anywhere shameem

Post on 05-Dec-2014

575 Views

Category:

Education

1 Downloads

Preview:

Click to see full reader

DESCRIPTION

 

TRANSCRIPT

A Market Anywhere

Professor Roy CampbellRyo Kawaguchi, Aukrit Unahalekhaka, Ang PanDepartment of Computer ScienceUniversity of Illinois at Urbana-Champaign

Presenter: Shameem Ahmed

• Easy access to market information can improve the quality of life [Journal of Economics, 2007]

Lack of Information

+ Integrated feature+ Simple+ Low cost

- 160 char per msg- No Visual Aids- Costly if …

Research GoalDesign a virtual market system– Applicable in developing countries– No costly infrastructure– Cell phone based– Use SMS– Reduce inherent problems of SMS system• Min # of transactions• Flexible inputs• Concise and well formatted output

+ Mobile & SMS- Menu Based

Restricted contentRestricted functionality

- Too many stepsCost patienceCost money

+ Mobile & SMS-Only market price Info- No NLP

Google SMS

+ Mobile & SMS- One Way Info- No NLP

Nokia Life Tools Farm Prices Trade Net

Current State of ArtCurrent State of Art

System OverviewSystem Overview

ZeepMobile

NLP EngineNLP Engine

Why?– SMS: Expressive and intuitive– Reduce # of msg– Little restriction on input format

How?– Feature extraction – Raw Text– Decision List

• Pattern and Field Type

Use CaseUse Case

Experimental Assessment

Compare to Cell Bazaar – 4 users (Unfamiliar with our NLP engine)– 100 queries in English– High field extraction success rate (over 90%)– 1 SMS (Cell bazaar needs 5)

Work In ProgressWork In Progress

• NLP for Thai language• Follow-up study in Thailand• Refine the design

Future WorkFuture Work

• Incorporate more languages• Extension to framework

ReferenceReference

• Cell Bazaar. www.cellbazaar.com • Zainudeen, A., Samarajiva, R., and Sivapragasam, N.

CellBazaar, a mobile-based e marketplace: Success factors and potential for expansion. ICA 2009.

• Nokia Life Tools. www.nokia.com/press/press-releases/showpressrelease?newsid=1266168

• Google SMS. www.google.com/sms

Backup SlidesBackup Slides

Algorithm NLPEngine (input)

1. while decision_list is not empty2. pop (pattern → field_type) from decision_list

3. if any phrase in input matches pattern4. fields[field_type] = normalized phrase5. remove phrase from input6. store fields and input in the database

NLP Engine AlgoNLP Engine Algo

Addressing Other Issues Addressing Other Issues

• Illiteracy –Basic education is must –Business opportunity for literate

• Misspelling /Missing words–Correct misspelling–Supplement missing words

• Unaware of market information• Product shortage • Excessive production• Middlemen• Not realize buyers and sellers match

Challenges Faced by Rural FarmersChallenges Faced by Rural Farmers

top related