Vis enkel innførsel

dc.contributor.advisorRamampiaro, Herindrasana
dc.contributor.authorThunestveit, Aina Elisabeth
dc.date.accessioned2016-10-04T14:00:43Z
dc.date.available2016-10-04T14:00:43Z
dc.date.created2016-06-15
dc.date.issued2016
dc.identifierntnudaim:14026
dc.identifier.urihttp://hdl.handle.net/11250/2412874
dc.description.abstractMany people use the Internet as a place for seeking opinions. With the increasing amount of user-based content on the web, there has been an emergence of research fields that uses sentiment analysis to take advantage of, and process this data. This could result in more satisfied customers, as relevant information will be easier to find. In this research, we present a new method, based on extracting and analyzing adjectives from user-based reviews. We exploit the idea that adjectives often contain a sentiment, and present a system entirely based on adjectives as sentiment deciders. Our experiments uses classical machine learning techniques like Random Forest and Support Vector Machines to predict the sentiment orientation of the reviews. Compared to baseline, our system improved results in all our experiments. Results also shows that with an accuracy of 94.7\%, our system performs better than state-of-the-art approach on a large dataset consisting of 50 000 movie reviews.
dc.languageeng
dc.publisherNTNU
dc.subjectInformatikk, Kunstig intelligens
dc.titleSentiment Analysis on User-Based Reviews: Movie Recommendation Case
dc.typeMaster thesis
dc.source.pagenumber100


Tilhørende fil(er)

Thumbnail
Thumbnail

Denne innførselen finnes i følgende samling(er)

Vis enkel innførsel