1 http://dbpedia.org 2 http://www.freebase.com 3 http://www.wikipedia.org Use of Semantic Co-relation in Target Audience Profiling Siaw Ling Lo, Dong Mei Shan, Viridis Liew School of Information Technology Nanyang Polytechnic Singapore {lo_siaw_ling, shan_dong_mei, viridis_liew}@nyp.edu.sg Abstract— With more companies doing businesses on social media, how can a company stand out from the increasingly crowded social space to find prospective customers from the audience in social media? It remains a challenge to sift through the huge amount of social media data, integrate the information and correlate among the different keywords or entities to form a more comprehensive view. The proposed solution aims to combine social media data and semantic linked data to extract relevant information and capture the relationship among the entities from content shared by the audience. With a targeted audience profiling, company is able to spend marketing dollars more effectively by sending the right offers to the right audience and hence maximize marketing efficiency and improve return of investment (ROI). Keywords-semantic web; linked data; target audience I. INTRODUCTION With the explosion of social media in recent years, social media giants such as Facebook, Twitter and LinkedIn have huge user-bases of over 1.7 billion with Facebook having 1 billion monthly active users [1]. More companies, in fact, 75% of them (according to statistics from BusinessWire) are leveraging on social media to do businesses. In addition, a survey [2] showed that majority of Fortune 100 companies are using more than one social media platforms to reach their potential audience. In the increasingly crowded social space, it is no longer feasible for a company to depend on gimmicks (such as incentive referrals) which may only provide short-term gain. While company can adopt approaches like mass marketing to all the fans or contacts available, the returns may not be justified by the effort and amount of money spent. Furthermore, there is a thin line between broadcasting general message and spamming so instead of attracting more audience, there is a high probability of losing current customer. Hence it makes sense to identify target audience to maximize the marketing efficiency and improve the ROI. Traditionally companies use mailers or emails to inform potential customer about their new products or promotions from third party listing or internal listing. With the proliferation of social media, companies are now using Facebook fan pages or Twitter to engage with their fans and followers or online audience. Currently there are generally two methods in identifying or reaching to audience on social media – keyword search and semantic tagging. While there are many advices or tips on how to find the target audience on social media, most of these concentrate on searching on profiles using key words related to product or brand. However while using this approach can retrieve lists of information, it is not capable to determine the relationship among the keywords and provide a more comprehensive view on the subject matter without the analysis of domain expert. Hence deciding which keywords to use may not be obvious to a non-expert and this may leads to inaccurate information extraction and hence a misunderstood market analysis. On top of this, there is a need to manually consolidate the list of profiles found and to ensure that the profiles match with the key words. Prior publications [3] [4] have proposed various approaches such as translating both social networks and semantic information into traditional Resource Description Framework (RDF) formats and use RDF methods for correlation, the use of semantic tagging to correlate the current social tagging approach respectively to make sense of the social media data. The two approaches mentioned required additional efforts of translating and tagging of current social media data which can be a daunting task considering the huge amount of data and the possible manual effort. Resources from LinkedData platform [5] such as DBpedia 1 , Freebase 2 , Wikipedia 3 have been used to make sense of the online sharing and content so that personalised services can be implemented to improve experience. For example, DBpedia is used to enhance on television experience with online social data [6]. The uses of semantic enrichment and temporal profile based on twitter content in recommending news for users [7]. Recently, Freebase, Wikipedia are used in predicting future news [8]. In this paper, we present a target audience profiling method using social media data correlated with semantic linked data. Freebase is used as the semantic source and Facebook data is used as the co-related social media source. There are two main modules of the method 1) subset identification of Facebook entities, and 2) similarity/relevance score calculation. This method provides companies a mean to identify potential customer from lists of audience on social media. With more comprehensive information on hand, companies are able to personalize or customize a product or service plan that will better suit the clusters of audience according to the profile identified.