Comparing the Arguments Within Social Networking Papers - Research Paper Example

Comparing The Arguments Within Social Networking Papers 1st 1st affiliation 1st line of address 2nd line of address Telephone number, incl. country code 1st authors E-mail address ABSTRACT This comparison paper provides an Executive Summary of three papers describing important aspects of the collation and interpretation of social networking data that will be of great interest to business enterprises currently conducting marketing and sales campaigns using social networking applications such as Facebook and LinkedIn - an increasingly common practice in the business world. The first paper describes the theoretical and mathematical work to produce a software application to implement a call-rating and feedback-driven application for producing social networking recommendations - a key business marketing tool. The second paper describes the implementation of a social network recommendation application, but in terms of a radically different model, described by the authors as the ‘homophily principle‘, which is expected to help improve any resulting recommendations’ accuracy and quality. The third paper describes the development of a social tagging system to provide social network users with an effective way to collaboratively annotate and organize items with specific tags. The authors describe a social tagging system as containing heterogeneous information including users’ tagging behaviors, social networks, tag semantics and item profiles. General Terms Management, Documentation, Human Factors. Keywords Social Networking, Applications, Recommendations, Tagging. 1. INTRODUCTION The first of the three papers to be considered is [3]. This paper describes the work of Yang, Steck and Liu in terms of both theory and mathematical work, carried out in the form of mathematical modelling, of what they call rating/feedback-driven recommender systems (RS). These systems constitute the very latest technology, so to speak, in the area of social networking, implementing both ‘Friends Circles’ and ‘Trust Circles’, which in turn serve to refine the concepts of ‘Friend’ and ‘Trust’, that have been part of social networking concepts such as Facebook for some years now. The objective of the paper is to demonstrate a quantitative method of implementing the concept of recommender systems as a software application. The purpose of this software application is to derive category-specific social trust circles from combining the available rating data with more general social network data. Yang, Steck and Liu have outlined several methods of weighting friends within circles based on their perceived expertise levels. Their experiments utilizing publicly available social network data serve to demonstrate that their circle-based recommendation models can better utilize a users social trust information, increasing recommendation accuracy - a key part of social networking, particularly if business objectives are to achieved. The second of the three papers to be considered is [2]. This paper describes the work of Shen and Jin on social network recommendation, but in terms of a radically different model, described by the authors as being derived from users linked within social networks tending to share a range of common interests and/or tastes (homophily principle), and by utilising this principle, help improve any resulting recommendations’ accuracy and quality. Shen and Jin also use mathematical models expressed as software applications to develop a joint Personal and Social Latent Factor (PSLF) model for social network recommendations which combines state-of-the-art collaborative filtering and the social network modeling approaches for social recommendation. Their PSLF model specifically extracts the social factor vectors for each user based on a current mixture membership stochastic block-model, which can explicitly express the whole range of social relationships. To complete their PSLF model, Shen and Jin have developed a scalable expectation-maximization (EM) algorithm to accelerate expectation calculations. As a benchmark, Shen and Jin compare their approach with two of the most statistically significant social recommendation data sets, namely Flixter and Douban. The Third Paper to be considered is [1]. The work of Feng and Wang describes the development of a social tagging system to provide social network users with an effective way to collaboratively annotate and organize items with specific tags. Feng and Wang describe a social tagging system as containing a variety of information including users’ tagging behaviors, social networks, tag semantics and item profiles. Employing a wide range of data like this will help alleviate the common ‘cold start’ problem caused by lack of immediately-available data. Feng and Wang model a social tagging system as a multi-type graph. In order to learn the weights of different types of nodes and edges, Feng and Wang propose an optimization framework, they call OptRank. They emphasize two key OptRank features. Firstly, edges and nodes are represented by features with different types of edges and nodes implying a different set of features. Secondly OptRank learns the best feature weights by maximizing the average AUC (Area Under the ROC Curve) of the tag recommender. Feng and Wang conducted their experiments using two publicly available social networking data sets : Delicious and Last.fm. Their results show that, firstly, OptRank outperforms existing graph based methods when only the relation is available. Secondly OptRank successfully improves results by incorporating social network, tag semantics and item profiles. 3. Common Themes All three papers deal with perceived problems with current social networking ‘sites’ such as Facebook and LinkedIn. To place these proposal in context, it must be borne in mind that social networking, that seems to have had it’s origins in the desire for young people to stay in contact and exchange information electronically, has now grown into a major marketing tool exploited by business enterprises at all levels. In order for business marketing to proceed satisfactorily, it is clearly necessary that the possibilities offered by ‘friends’ and ‘recommendations’ be exploited to the full in order for enterprises to ‘sell’ their products or benefits (which is what ‘service industries’ are about, after all!) to potential customers. The software applications offered by the authors of all three papers are clearly an excellent means of achieving this aim, and offer facilities more traditionally offered by advertising agencies, market research bodies and product statistics databases, but at a small fraction of the cost. This makes is practical for even the smallest and most ‘niche’ enterprise to levels of enjoy marketing and advertising opportunities that, until the advent of social networking, were only available to the largest corporations and others (such as governments) that were able to maintain their own advertising, marketing, public relations and statistics agencies. Examining all three papers has shown that all the software applications have been designed to be relatively straightforward to implement on a commercial scale, and hence constitute products that ‘mine’ the customer data already present within social networking systems and databases, collate and interpret this data, and present the results to the ‘owner’ of the application with the data appropriate to that owner’s business enterprise. 4. Discordant Themes The first two papers deal with social network recommendations, and the third deals with ‘tagging’ social network data. It may therefore be said that the three different papers deal with two different themes, and the first two papers examined deal with the theme of social networking ‘recommendations’ in different ways. It must also be said that the second and third papers examined have used available social networking data to test their software applications, whereas the first has not. Therefore, it may be put forward that the second and third papers will, when the results of the software application ‘runs’ are examined in detail, produce more valid results, because they have been based around ‘live’ social networking data sets. However, it appears that there are no discordant themes as such, as the term ‘discordant’ implies that there are two or more ways being presented of achieving the same goal, and that those ways do not agree. Examining the three papers in question shows instead three different objectives, which implies that the software applications may be made to work effectively together and produce a much better capture, collation and implementation of the available social networking data than one application working on a ‘stand alone’ basis. 5. Non-Overlapping Themes Given that the three papers being examined provide three different software applications, it may be therefore be said that all three themes of the papers do not overlap, but may nonetheless be made to work together in terms of software applications, given a common social networking database and some requirements on how to use the collated and interpreted social networking data. Firstly, the work of Yang, Steck and Liu shows that the best subset of a users friends, that is to say, an inferred circle, corresponds to the best recommendations in a given category of interest. As such inferred circles are tailored towards various social networking data categories, they will often from explicit circles of friends that have recently become popular in online social networks. Yang, Steck and Liu have proposed methods for inferring category-specific circles, and hence assign weights to the friends within each circle. In their experiments on publicly available data, the work of Yang, Steck and Liu showed significant improvements over existing approaches that use mixed social network information. Secondly, the work of Shen and Jin has led to the development of a a joint personal and social latent factor (PSLF) model which combines the collaborative filtering and social network modeling approaches for social recommendation. Their approach appears to have shown to have better prediction accuracy over existing social network data capture, collation and analysis approaches, which will be of great interest to enterprises at all levels. Thirdly, the work of Feng and Wang has led to an optimization-based graph method for personalized social networking data tag recommendation. To maximize the data set different sources of information are incorporated into the optimization framework. Exactly how to utilize social networking ‘tagging’ in this way needs to be further investigated from a business perspective. 6. Conclusion The three papers examined have illustrated excellent examples of the work currently underway to utilize social networking data for business purposes. All three papers present software applications which, when used together with a common social networking database, will provide business enterprises at all levels with a much more fine-grained interpretation of social networking data, leading to much more focused marketing efforts as part of sales campaigns mounted using social networking, a practice increasingly used by the business community as a whole. 7. References [1] Incorporating Heterogeneous Information for Personalized Tag Recommendation in Social Tagging Systems, Wei Feng, Tsinghua University, Beijing, China, Jianyong Wang, Tsinghua University, Beijing, China. August 2012. Association of Computing Machinery. Supplied With Assignment [2] Learning Personal & Social Latent Factor Model for Social Recommendation. Yelong Shen, Department of Computer Science, Kent State University, Ruoming Jin, Department of Computer Science Kent State University, August 2012. Association of Computing Machinery. Supplied With Assignment [3] Circle-based Recommendation in Online Social Networks, Xiwang Yang, ECE Department, Polytechnic Institute of NYU, Brooklyn, New York, Harald Steck, Bell Labs, Alcatel-Lucent, Murray Hill, New Jersey, Yong Liu, ECE Department, Polytechnic Institute of NYU, Brooklyn, New York. n.d. Association of Computing Machinery. Supplied With Assignment Read More