Abstract:
Deforestation, pollution of water bodies, natural disasters are few environmental
problems in Sri Lanka. Communities who suffer from environmental problems are
not attaining solutions due to lack of awareness and carelessness in agreeable
parties. In most influential situations, community uses mass media for gaining
responsiveness of authorized parties. Virtual communities in Sri Lanka used social
media for emphasizing numerou s forms of social problems occurred within last
decade. The intention of this research is to assist amenable organizations in making
better decisions according to public views.
The application is obtaining the content of environmental problems with
crowd sourcing . The virtual community can report environmental problems by
using text and images. Users of the application can vote or comment on the
problems reported by others. Each problem will receive points according to a pre
defined algorithm based on the number of up or down votes and polarity of user
comments they have received. The application highlights genuine high quality
problems while allocating points for each user. Topic based text categorization
techniques which used in this prototype is effectiv e for filtering out environmental
related information which is not used yet for the most important natural language
applications. The polarity of user comments is generated as positive or negative by
using sentiment analysis tools.
The application is provi
ding 90% of accuracy in highlighting environmental
problems by considering pre defined user features, user activity measures and
community process features. The accuracy of the application can be automatically
improved with the growth of corpus provided by the users of the application.