with Java and Lenskit framework. The study resulted in two real data sets, namely MovieLens and. TripAdvisor, in which the authors have shown positive results.
International Journal of Knowledge Society Research Volume 7 • Issue 3 • July-September 2016
Improving Collaborative Filtering Algorithms:
Sentiment-based Approach in Social Network Firas Ben Kharrat, IHEC, University of Carthage, Carthage, Tunisia Aymen Elkhleifi, LaLIC, Paris-Sorbonne University, Paris, France Rim Faiz, IHEC, University of Carthage, Carthage, Tunisia
ABSTRACT This paper puts forward a new recommendation algorithm based on semantic analysis as well as new measurements. Like Facebook, Social network is considered as one of the most well-prominent Web 2.0 applications and relevant services elaborating into functional ways for sharing opinions. Thereupon, social network web sites have since become valuable data sources for opinion mining. This paper proposes to introduce an external resource a sentiment from comments posted by users in order to anticipate recommendation and also to lessen the cold-start problem. The originality of the suggested approach means that posts are not merely characterized by an opinion score, but receive an opinion grade notion in the post instead. In general, the authors’ approach has been implemented with Java and Lenskit framework. The study resulted in two real data sets, namely MovieLens and TripAdvisor, in which the authors have shown positive results. They compared their algorithm to SVD and Slope One algorithms. They have fulfilled an amelioration of 10% in precision and recall along with an improvement of 12% in RMSE and nDCG. Keywords Recommendation System, Semantic Analysis, Social Network Analysis, Social Recommendation, User Cold Start
INTRODUCTION Nowadays, the continuous progress of data and communications technologies, the increasing use of digital technologies among the people and the development of the users’ lifestyles and norms necessitate innovative types of systems. The success of these systems will rely upon their capability to operate as a unit, to supply services to their users in every place and at any moment as well as to accustom themselves to the changes surrounding their environments. Recommender systems help users to identify their interests and sets of choices by predicting the invalidity of an item or a group of items to these users. They are defined as a special type of information filtering that provide information about which items may be interesting to users. Recommender systems generate a personalized recommendation for users based on a set of previously rating items. However, according to (Baldoni et al., 2011) the social web principles, social network and platforms promote the participation of users in many ways, stimulating the expression of opinions about the contents inserted by other users, by ‘Like’, ‘Share’ annotations, star-rating systems... This huge amount of data is a precious information source about perceptions, trends and feelings, and a lot of research is being carried on to identify ways for extracting meaningful information from these data. In fact, the social network could be considered an important source of recommendation. This paper endeavors to combine two types of DOI: 10.4018/IJKSR.2016070107 Copyright © 2016, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
99
International Journal of Knowledge Society Research Volume 7 • Issue 3 • July-September 2016
information respectively called social and semantic information. The main purpose is to compute recommendation lists through making a visit to a limited part of social network. We have improved the Slope One and SVD algorithms by external knowledge and through some experiment results. The approach has been applied in Java and tested on the MovieLens and TripAdvisor data sets. The results demonstrate that the additional knowledge achieved by the semantic analysis contributes to the amelioration of the quality recommendation in terms of accuracy. Hence, this study foregrounds two major issues. The first part seeks to gather as well as parse data, while the second one attempts to use external resources to come up with relevant suggestions. We construct this paper as following: We start with some major research literature on recommendation systems. Next, we highlight our approach via applying user’s sentiment in order to raise the value of recommendation algorithm. We clarify the experiment immediately followed by an estimation of our algorithm. In conclusion, we briefly summarize our proposed study providing a few remarks on future works. EASE OF USE According to Burke (Burke, 2002), Recommender systems have three fundamental categories: collaborative-filtering, content-based and hybrid. The hybrid recommender system combines the content-based and the collaborative-filtering regarding the kind of advocated data Recommendation types are explained in details with their limitations in this section: Content-based: It advocates items that are akin to those the users selected on time before (Adomavicius & Tuzhilin, 2005; Burke, 2002). Items like products, services or even people are identified by their related characteristics. User’s favorite choices, which are preserved in the user’s profile and taking into account those associated criteria, emerge in already estimated items. According to (Schafer et al., 1999) content-based technique is termed item-to–item affinity. We show some usual research on which these traits are revolved around this technique, which are: ◦◦ NewsWeeder (Lang, 1995) is newsgroup, an internet base-discussion, filtering system. It charges users to rank some not viewed news articles according to their preferences. The applications were built on schematic tree-shaped diagram, sensory network and graphics based on vector. ◦◦ Pazzani et al., 1997 suggests a system able to bring forth web sites based on issues that the users have to be involved in regarding to their typical interest. In this experimentation, the applied system is Bayesian classifier. ◦◦ Zhang et al., 2002 suggests to differentiate in the midst of queries appropriate files including new data and using Bayesian system. Collaborative Filtering: CF is also termed people-to-people correlation. ◦◦ Tapestry (Goldberg et al., 1992) extracts electronic script on a drafted issue by a particular user. ◦◦ Ringo (Shardanand & Maes, 1995) advocated recording of music and artists established by resemblances among user’s tastes as well as others. ◦◦ Hoaks (Hill & Terveen, 1996; Terveen et al., 1997) proposes home pages from news messages. If followers are interested, the system will not only find the person who got in a touch but also identify the web page. ◦◦ GroupLens (Konstan et al., 1997; Resnick et al., 1994) makes an advance in a system that deals with extracting data that assesses newsgroups article. ◦◦ Jester (Goldberg et al., 2001) is an online funny recommending the system. It uses a Collaborative Filtering algorithm termed Eigentaste and it is adjacent to algorithm for the disconnected phase. Next, the author utilizes a repetitive rectangular classification method for the offline stage. 100
18 more pages are available in the full version of this document, which may be purchased using the "Add to Cart" button on the product's webpage: www.igi-global.com/article/improving-collaborative-filteringalgorithms/170506?camid=4v1
This title is available in InfoSci-Journals, InfoSci-Journal Disciplines Communications and Social Science, InfoSciCivic Engagement, Sustainable Planning, and Crisis Response eJournal Collection, InfoSci-Knowledge Discovery, Information Management, and Storage eJournal Collection. Recommend this product to your librarian: www.igi-global.com/e-resources/libraryrecommendation/?id=2
Related Content Delivering Educational Games to Mobile Devices Miroslav Minovic, Miloš Milovanovic and Dušan Starcevic (2011). International Journal of Knowledge Society Research (pp. 47-58).
www.igi-global.com/article/delivering-educational-games-mobiledevices/55264?camid=4v1a Manufacturing Knowledge: Technology, Culture, and Social Inequality at Work Steven Vallas (2000). Social Dimensions of Information Technology: Issues for the New Millennium (pp. 236-254).
www.igi-global.com/chapter/manufacturing-knowledge-technology-culturesocial/29120?camid=4v1a Sporting Safe in the Liminal Sphere: “Tactics” and Facebook Santosh Khadka (2014). Emerging Pedagogies in the Networked Knowledge Society: Practices Integrating Social Media and Globalization (pp. 206-216).
www.igi-global.com/chapter/sporting-safe-in-the-liminalsphere/96062?camid=4v1a
A E-Business Case of Study: Modelling the Quality of the Wine using its Physicochemical and Qualitative Properties Maria Vargas-Vera, Camilo Salles, Joaquin Parot and Sebastian Letelier (2017). International Journal of Knowledge Society Research (pp. 1-20).
www.igi-global.com/article/a-e-business-case-of-study/183678?camid=4v1a
