Computer Engineering / Bilgisayar Mühendisliği

Permanent URI for this collectionhttps://hdl.handle.net/11147/10

Browse

Filters

2010 - 20123

Settings

Author: search.filters.author.Tapucu, DilekScopus Q: search.filters.scopusQuartile.Q4

Search Results

Now showing 1 - 3 of 3
  • Conference Object
    A Roadmap for Semantifying Recommender Systems Using Preference Management
    (Springer, 2010) Tapucu, Dilek; Tekbacak, Fatih; Ünalır, Murat Osman; Kasap, Seda
    The work developed in this paper presents an innovative solution in the field of recommender systems. Our aim is to create integration architecture for improving recommendation effectiveness that obtains user preferences found implicitly in domain knowledge. This approach is divided into four steps. The first step is based on semantifying domain knowledge. In this step, domain ontology will be analyzed. The second step is to define an innovative hybrid recommendation algorithm based upon collaborative filtering and content filtering. The third step is based on preference modeling approach. And in the fourth step preference model and recommendation algorithm will be integrated. Finally, this work will be realized on Netflix movie data source. © 2011 Springer Science+Business Media B.V.
  • Conference Object
    Citation - Scopus: 5
    Performance Comparison of Combined Collaborative Filtering Algorithms for Recommender Systems
    (Institute of Electrical and Electronics Engineers Inc., 2012) Tapucu, Dilek; Kasap, Seda; Tekbacak, Fatih
    Recommender systems have a goal to make personalized recommendations by using filtering algorithms. Collaborative filtering (CF) is one of the most popular techniques for recommender systems. As usual, huge number of the datasets on the Internet increase the amount of time to work on data. This challenge enforces people to improve better algorithms for processing data with user preferences and recommending the most appropriate item to the users. In this paper, we analyze CF algorithms and present results for combined user-based/item-based CF algorithms for different size of datasets. Our goal is to show combined solution results using Loglikelihood, Spearman, Tanimoto and Pearson algorithms. The contribution is to describe which user based CF algorithms and user/item based combined CF algorithms perform better according to dataset, sparsity, execution time and k-neighborhood values. © 2012 IEEE.
  • Conference Object
    Citation - Scopus: 18
    New Features for Sentiment Analysis: Do Sentences Matter?
    (CEUR Workshop Proceedings, 2012) Gezici, Gizem; Yanıkoğlu, Berrin; Tapucu, Dilek; Saygın, Yücel
    In this work, we propose and evaluate new features to be used in a word polarity based approach to sentiment classification. In particular, we analyze sentences as the first step before estimating the overall review polarity. We consider different aspects of sentences, such as length, purity, irrealis content, subjectivity, and position within the opinionated text. This analysis is then used to find sentences that may convey better information about the overall review polarity. The TripAdvisor dataset is used to evaluate the effect of sentence level features on polarity classification. Our initial results indicate a small improvement in classification accuracy when using the newly proposed features. However, the benefit of these features is not limited to improving sentiment classification accuracy since sentence level features can be used for other important tasks such as review summarization.