Sentiment Web Mining Architecture

Shahriar Movafaghi, Jack Bullock

Southern New Hampshire University

In this paper we discuss the architecture of the system under development

to capture the sentiment of web users regarding any topic such as retail
products, financial instruments (FI), or social issues like immigration. The
first step is knowledge acquisition. A Sentiment Web Mining (SWM)
system requires acquisition of knowledge from several sources on the
web. Such knowledge may be found on blogs, social networks, emails, or
news. A SWM system has customization and personalization capabilities.
For our purposes, customization occurs when the SWM user can change
his/her preferences to select specific sites to be used for data mining and
evaluation. Personalization occurs when the system decides which sites to
be used for data mining based on the user profile. The user profile
dynamically changes depending on the type of user request from the
system and the specific sites the user visits to verify the result of the SWM
system.

The second step is knowledge storage which involves the creation of a

database. Appropriate web sites will be indexed and tagged. Taxonomy is
the hardest part of this step. In this paper we will demonstrate a unique
way of tagging the knowledge obtained from the web. The third step is the
knowledge analysis/data mining. A SWM system will use a series of off –
the- shelf knowledge analysis/data mining tools including SWM knowledge
analysis/data mining engine which is based on web services technology.
The type of question used can be: 1) The volume of sentiment for
particular topic; 2) The intensity of sentiment (good or bad) for a particular
topic; 3) the interrelationship between the writers of material written on the
web, especially if the writer is anonymous; 4) Who is the leader(s) of the
sentiment? If the information is maliciously posted on the web the user
may want to pursue it according to the law.

The last step is dissemination of knowledge to the user(s). A SWM system

uses third party visualization tools as well as web based user interfaces
and reports that are written internally. The presentation component of
SWM system is decoupled from other components, namely, process
1
component, business rule component, and data access component for
ease of maintainability.

Shahriar Movafaghi s.movafaghi@snhu.edu +1 603 557 4189

Jack Bullock jdbullock.2@gmail.com +1 207-385-3711

Download copies of COINs 2009 research and industry papers at

ScienceDirect.

Link: http://www.sciencedirect.com/science/issue/59087-2010-999979995-
2182758

Procedia - Social and Behavioral Sciences, Volume 2, Issue 4, The 1st

Collaborative Innovation Networks Conference - COINs2009. Edited by
Kenneth Riopelle, Peter Gloor, Christine Miller and Julia Gluesing.

Connect to the COINs 2010 Conference community across these media

platforms:

o COINs 2010 http://www.coins2010.com

o Facebook http://www.facebook.com/pages/Savannah-
GA/Collaborative-Innovation-Networks-COINS2010-
Conference/102489653133049
o Flickr http://www.flickr.com/photos/coinsconference/
o Livestream http://www.livestream.com/coinsconference
o Scribd http://www.scribd.com/SwarmCreativity
o Slideshare http://www.slideshare.net/IOpen2
o Twitter http://twitter.com/coins_2010
o Hashtag #COINS2010
o Vimeo http://www.vimeo.com/user4147060
o You Tube http://www.youtube.com/coinsconference

The COINs 2010 conference, Oct. 7–9, 2010, is presented by I-Open and
the COINs Collaborative, an initiative of the Savannah College of Art and
Design, Wayne State University College of Engineering Department of
Industrial and Manufacturing Engineering, and the Massachusetts Institute
of Technology Center for Collective Intelligence. The collaborative builds
open knowledge networks to advance the emerging science of
collaboration for research and industry competitive advantage. Hosted by
SCAD. For more information about the COINs 2010 conference, visit
www.coins2010.com.

2
3