IBIMA Publishing Journal of Internet Social Networking & Virtual Communities http://www.ibimapublishing.com/journals/JISNVC/jisnvc.html Vol. 2011 (2011), Article ID 685060, 14 pages DOI: 10.5171/2011.685060

The Information Potential and Temporal Elements of Web 2.0 Folksonomy for User Profiling in Personalized Information Retrieval

Wern Han Lim, Saadat M. Alhashmi and Eu-Gene Siew

Monash University, Kuala Lumpur, Malaysia ______

Abstract

The World Wide Web (WWW) has grown into a rich repository of information. The vast amount of data on the web however would require the users to be skilled in querying generic information retrieval (GIR) systems to meet their information need. This is further complicated by various noises on the web such as spam and advertisement. Personalized information retrieval (PIR) systems have the potential to meet users’ information need effectively and efficiently. This paper focuses on the user profiling (UP) aspect of PIR systems as it would directly determine the system’s ability to address the users’ information need. With the high adoption of web 2.0 systems among the users of the web, web 2.0 systems proved to be an important source of user information for improvement in user profiling. One of such systems is collaborative tagging systems otherwise known as folksonomy. This paper explores the information potential of folksonomy systems in improving user profilers. A case study of the Delicious social bookmarking system was conducted to explore temporal elements of folksonomy which is neglected in previous approaches to improve the performance of user profilers. We conclude that folksonomy systems have the information potential to enhance the performance of user profilers.

Keywords: Information Retrieval, Personalization, User Profiling, Folksonomy, Temporal, Delicious ______

Introduction well as some knowledge in the context of their information need. Thus, the Users constantly find themselves in need for performance of the GIR systems in meeting information in today’s world; the World the users’ information need would vary Wide Web (WWW) is the biggest repository depending on the users and their expertise. of information available to the public. The vast amount of information on the WWW Such circumstances motivate researches to however makes the retrieval of particular look for an alternative to GIR systems - information in the need of the users difficult. personalized information retrieval (PIR) Thus, generic information retrieval (GIR) systems. Unlike GIR systems, PIR systems systems have been developed to assist the would try and provide the information that users. GIR systems were extremely successful the users need without asking for it in meeting the users’ information need. The implicitly. This is done through the PIR ever increasing amount of information system’s knowledge about the users as well coupled with the high amount of noise on the as the ability to determine the context of the web however would require the users to users’ information need. Thus, the have IR skills in obtaining the required performance of PIR systems is less information such as query formulation as dependent on the expertise of the users

Copyright © 2011 Wern Han Lim, Saadat M. Alhashmi and Eu-Gene Siew. This is an open access article distributed under the Creative Commons Attribution License unported 3.0, which permits unrestricted use, distribution, and reproduction in any medium, provided that original work is properly cited. Contact author: Wern Han Lim. E-mail: [email protected] Journal of Internet Social Networking & Virtual Communities 2

while meeting the users’ information need focuses on the folksonomy component of effectively and efficiently. Besides that, the IR web 2.0 systems. Our study into this topic is system must also deal with the potential supported by a case study into the Delicious vocabulary problem such as homonyms and social bookmarking system where we synonyms. showcase the existence of temporal properties for folksonomy (section 3.3). The For example, a user with low expertise in IR temporal aspect of folksonomy is often could produce a query with the keyword neglected by past researchers and we look ‘jaguar’ when he/ she as a science student is into the information potential from such looking for information about the animal. A elements. We summarize our findings and GIR system would return results of both the some applications for the improved user animal as well as the automotive proiling in section 5. manufacturer. In return, the user would then need to look through the results for the Personalized Information Retrieval (PIR) information that he needs. Such situation would not occur within a PIR system. The PIR Most of the information retrieval (IR) system would know that the user himself is a systems of the current age such as search science student and the context of the ‘jaguar’ engines (Yahoo!, Google etc.) are generic keyword is in that of the animal and not the (GIR) – for a given query or request, the same automotive manufacturer. Thus, the PIR results would be returned regardless of the system would return the suitable results for context or users’ interest. An example would the users despite the ambiguous query given be the jaguar analogy given above in the by the user. introduction. While the results may change overtime according to the algorithm, the To do so, the PIR system must have results returned are often generic of various knowledge about the users – data about the contexts. There is no doubt that GIR systems users collected and inferred into information are successful over the years in meeting the stored via the PIR system’s user profiler. The information need of the users. The efficiency best source of information about the users of GIR systems is however decreasing with would come explicitly from the users the increasing amount of information and themselves but studies found that users are resources on the web. The diversity of such unwilling to provide information about resources together with the associated themselves. Thus, implicit collection of user vocabulary problem in natural language data is the choice of most PIR systems and to processing (NLP) would require the users to do so, the PIR system must have reliable formulate suitable queries to meet their sources of user information. This has been a information need. Search query is usually an concern in various researches and with the approximation of the user’s information need advent of web 2.0; PIR systems have a due to ambiguity (vocabulary problem), credible alternative source of user mismatches as well as being dependent on information. The interaction between users the user’s vocabulary and knowledge (Wang and web 2.0 systems would provide valuable and Davison, 2008). Thus, a personalized information about the users to be used in the information retrieval (PIR) system is web2PIR’s user profiling. proposed as the solution. While users are accustomed to GIR, it is found that more than In this paper, we introduce the 80% of users would prefer to receive such personalization of IR systems over generic personalized results over generic ones (Lee systems to meet the information need of the et al., 2005). users in section 2. We then continue on the addition of web 2.0 systems into PIR systems The objective of PIR systems is to provide and how web 2.0 could improve user users with the information that they want or proiling in section 3 and 4. The paper need without asking them explicitly 3 Journal of Internet Social Networking & Virtual Communities

(Mulvenna et al., 2000). Past researchers The Web 2.0 Approach have conducted studies into this and found PIR systems to outperform GIR systems The web 2.0 is the next generation of the various approaches: - World Wide Web (WWW). It is an interactive platform for information sharing and - Re-ranking of results (Daoud et al., 2009, collaboration centered on the users known as Haveliwala, 2002) user generated content and not limited to the authors as the generation before. Web 2.0 - Recommendation of information platforms include Wikipedia, social tagging (Mobasher et al., 2000, Forsati et al., 2009) (Folksonomy), blogs and social networks. Information on web 2.0 is constantly updated - Query enhancement/ reformulation/ at a high rate with the interaction of users. recommendation (Liu et al., 2004, Sieg et Researchers in their studies discover and al., 2004) acknowledge web 2.0 platforms as a valuable source of information and their potential to Traditional PIR systems as stated above rely be used within information systems (Marlow on direct user interaction to learn about the et al., 2006) especially in user profiling. users such as the resources which are viewed, saved, bookmarked and bookmarked User Profiling with Folksonomy by the users especially on the client side. Thus, the PIR system would need to process Collaborative tagging system or otherwise these resources for to infer the interest of the known as a folksonomy system is a system users. Such resources lack valuable metadata which allows the users and web communities as they are only stated by the authors and to attach terms, keywords, tags or thus would require classification of the annotations to shared resources (Golder and resources and the content itself to infer the Huberman, 2005). Such attachments are interest of the users for user profiling. Due to short/ brief descriptions which sum up the the dynamic nature of web resources, associated resources (Marlow et al., 2006, classification of such resources is often Heymann et al., 2008) and help to classify as complex with low precision. Such scenario well as organize resources (Wal, 2005) motivates the research into an alternative within their platform. Such addition to source for user information with reliable resources on the WWW would enhance the information and easy processing for user metadata of resources which are beneficial to profiling – web 2.0. The improvement in user information systems (IS). profiling would directly impact the performance of PIR systems.

Journal of Internet Social Networking & Virtual Communities 4

Fig 1. The Folksonomy Model

The basic folksonomy model is built from the Studies found high overlapping of terms 3 main elements: the users of folksonomy between expert reviews and annotations - systems, the resources on folksonomy 73.01% term overlapping between the systems and the annotations/ tags which are content of expert reviews with tags of music used by the users to describe/ organize resources (Last.FM) (Bischoff et al., 2008). folksonomy resources (see igure 1). Thus, The same observation is made for other the model could be viewed as a tripartite resource formats such as the bookmarked graph with the 3 elements with annotations webpages of Delicious. Annotations are connecting the users with resources (Mika, found to have a high overlapping of terms 2005, Marlow et al., 2006). with both the title and content of resources (Heymann et al., 2008). We could conclude Folksonomy Annotations that annotations are highly beneficial in describing the resource while providing In general, studies found that the usage of additional evidence of resource popularity annotations is varied but highly popular, (Amitay et al., 2009). frequent and stable (Golder and Huberman, 2005). Among the known usage of Annotations could be used directly as queries annotations include: - due to their high overlapping with queries (Bischoff et al., 2008, Amitay et al., 2009). - The classification of resource type and is Beside the semantic meanings which directly used for the personal organization annotations carry, annotations are also found of resources. Classification using to be a potential measure of the popularity of annotations were found to successfully resources as a form of rating/ voting for replace the use of traditional tagged resources (Liang et al., 2008). As categorization of classes such as ODP (Xu annotations are assigned by the users, it et al., 2008) for PIR due to their dynamic could reflect the interest of the users more nature and larger coverage. accurately especially in semantics (Han et al., 2010). - Self noting of the users on their perception or rather a personal metadata about the Thus, social annotations provide accurate resources (including quality) information about the users to be used for 5 Journal of Internet Social Networking & Virtual Communities

user profiling especially when the Relation between Annotations annotations are assigned by the users themselves. There is no need for much extra A tag cloud could be represented as a vector processing of annotations as they are usually of annotations weighted according to the short and precise unlike full text methods. frequency of usage by the user (Szomszor et Thus, user profiling using annotations are al., 2007, Noll and Meinel, 2007). The effective and efficient. information provided by tag clouds (and their direct usage as user profile) are Resource Annotations however insufficient due to the following reasons (Michlmayr and Cayzer, 2007): - Traditional PIR systems perform user - Tag clouds do not maintain the profiling based on the annotations used by relationship between annotations which the users directly (Ha et al., 2007, Diederich hold valuable semantic value and context and Iofciu, 2006) known as ‘user annotations’ in the understanding of the users and their (UA) and not the other annotations which are interest (Milicevic et al., 2010) especially in associated with the related resources, the dealing with the vocabulary problem ‘resource annotations’ (RA). This is a concept especially synonyms and homonyms. explored by the authors who observed that resources could be used to indicate the - The popular annotations which are highly interest of the users as users do not only use weighted within tag clouds are usually very the annotations which they assigned but also general; thus the specificity of the user the other annotations of the assigned interest. resources (Cai et al., 2010). Annotation frequency is highly explored in the past to - Does not take into account the temporal obtain the interest of the users (Noll and element associated with bookmarked Meinel, 2007). The addition of annotations resources and annotations. from related resources improves the performance of PIR systems as opposed to - Resources might not contain matching non-personalized methods, using only user annotations with the user profile but are of annotations alone or using only resource interest to the user. Thus, tag cloud alone is annotations alone (Xu et al., 2008). This insufficient as the user profile (but rather could be partly explained by the additional need to be used to obtain the interest of the information (annotations) which helps the user). PIR system to establish the context and classification of resources. PIR systems based upon folksonomy systems should always maintain the semantic relation The addition of resource annotations allow between annotations. If two annotations are the PIR systems to profile users who have used in combination with each other (thus interacted with those resources and not only co-occur) by a certain user, there is some users who only contributed (via user kind of semantic relationship between them annotations) within folksonomy systems. (Michlmayr et al., 2007). The more often Besides that, these resources could very well annotations are used in association with each appear in other platforms such as social other, the stronger the relationship is. Such networks which could be interacted by other relationship would make sense to the users and not only via folksonomy systems. annotator and thus is not community driven This would allow a wider range of users to which is usable by PIR systems for user enjoy the benefits of personalized proiling. It is found that 92% of resources information retrieval. within Delicious are annotated with at least 2 annotations (Michlmayr and Cayzer, 2007).

Journal of Internet Social Networking & Virtual Communities 6

The relationship between the annotations known as the user’s personomy (Hotho et al., could also be used to obtain the context of 2006). In general, these are the following the annotations in obtaining the user’s ways to obtain related annotations of the interest. Without the relation between users according to user action in associating annotations, this would not be possible due with these resources: - to the vocabulary problem associated with terms/ keywords such as synonym (group of - Annotations added manually by the users terms with the same meaning) and to their profile. homonyms (same spelling of terms with different meaning). - Annotations used directly by the users.

Temporal Element of Annotations - Annotations from resources related to the users through users’ actions (browsing, Besides that, annotations do display viewing, saving, bookmarking, tagging, temporal properties as shown within the sharing etc). adaptive approach such as the Add-A-Tag algorithm (Michlmayr and Cayzer, 2007). Traditionally, the simplest approach is to Within such approach, the PIR systems take maintain the annotations in their original into account the fact that bookmarks (and form without any processing within the user indirectly annotations) have age and thus profile. Such approach, or otherwise known signifies the changes in user interest. With as the naïve approach, models the user this in mind, the user profile should be profile as a weighted vector of tags/ updated overtime. Under such approach, the annotations (Szomszor et al., 2007, Noll and users’ annotations could be treated as a Meinel, 2007) with the weights determined continuous stream of information and thus by the frequency of the annotation within the sequence could be used to update the user user’s personomy. The naïve approach in profile. The user profile could be updated in modeling the user profile is simple, fast and various ways such as via the ant algorithm could be used to determine the suitable (an extension on the evaporation technique) resources by comparing the cosine similarity where the edges would degrade overtime of the vector between the user and the according to a certain percentage. resource (Diederich and Iofciu, 2006).

The temporal properties of annotations are The naïve approach would be suitable for essential in user profiling to detect the latter users with only a single interest which is interest of the user through newer however highly unlikely from the study in the combination of annotations and bookmarks distribution of annotations within the which are currently more important to the Delicious social bookmarking system (Au et user. In their user study (Michlmayr and al., 2008). In that study, the authors conclude Cayzer, 2007) in comparing the co- that users of Delicious have multiple occurrence approach with the add-a-tag interests and thus proposed to obtain the approach, the authors found that users are interest of the users through clustering of fond of the performance of both approaches resources into multiple annotation vectors valuing the long term relationship between and thus obtain the users’ interests as topics. annotations as well as the ability of the Add- A-Tag algorithm to adapt to recent changes. Besides that, the naïve approach does not take into account the relationship between User Modeling with Folksonomy annotations (Milicevic et al., 2010); and thus the context of the annotation which is vital The information potential of annotations when dealing with the vocabulary problem could be used for user profiling. The especially synonyms and homonyms. If two annotations associated with the users are annotations are used in combination with 7 Journal of Internet Social Networking & Virtual Communities

each other (thus co-occur) by a certain user, through annotations. Thus, we attempt to there is some kind of semantic relationship study the temporal effect of folksonomy between them (Michlmayr et al., 2007). A systems through a case study of Delicious simple solution to this would be to maintain (see igure 2 to igure 5). the user profile as multiple tag clouds (each from the associated resource) (Szomszor et These are the following hypotheses which we al., 2007) or weighted graph (Michlmayr et would like to assure from our case study: - al., 2007). The weighted graph approach has the advantage in terms of updating the user - Hypothesis 1 (H1) : Folksonomy resources profile due to the temporal element which do exhibit temporal properties with would affect the interest of the users. varying lifespan based upon the interaction of the users Temporal Effect of Folksonomy Resources - Hypothesis 2 (H2) : Folksonomy resources Unlike traditional web resources however, are enhanced overtime through users resources within web 2.0 systems are interaction instead of only on the first day enhanced overtime through the users’ where bookmarks are added and shared on interaction showcasing the temporal Delicious properties of folksonomy systems. This is often overlooked or unexplored by We obtained 32401 unique bookmarks from researchers where past researches only take our crawling of Delicious for our case study. into account the information on folksonomy From the data extracted, we found that only resources only at the point of interaction. 4.7% of the Delicious bookmarks have a Any information after the point of interaction lifespan of only 1 during. This exhibits the by the users is not taken into consideration temporal element of folksonomy resources for user profiling. As folksonomy resources where a big majority of bookmarks on exhibit temporal properties which we would Delicious are still being interacted and prove via our case study, the metadata of the updated by users over time. As stated before, resources increases; which would provide past PIR systems would only extract the valuable information about the users which information at the point of interaction. Thus, interacted with them. Such information users who interacted with the resources on include additional annotations (might the first day would miss out on the potential include temporal based information such as valuable information (to establish context trends etc), new relations between and overcome the vocabulary problem) from annotations as well as user comments/ notes. the resources with lifespan of more than one This creates an information gap between the day. We found that the Delicious bookmarks user’s current point of interaction have a median of 9 days which means that (information up to this point is being used for around 50% of the bookmarks have a user profiling) and subsequent interaction lifespan of more than 9 unique days. Our points of other users (new information added findings support the first hypothesis on by the other users). As such, such folksonomy resources having temporal information should be used to update the properties. profile of the users as the resources are being updated. To support our 2 nd hypothesis, we attempt to show that folksonomy resources do receive Case Study: Delicious new unique annotations (we define unique annotations as annotations which terms had Delicious is one of the most successful not appeared in earlier days for the same collaborative tagging systems on the WWW bookmark) throughout their lifespan. We for web resources allowing users to observe that the number of unique bookmark web resources and organize them annotations decreases as the age of the Journal of Internet Social Networking & Virtual Communities 8

bookmark increases. An interesting potentially be valuable information which observation would be on the average number could be used to establish the context of the of unique annotations (empty ones are not resources as well as to overcome the taken into consideration) where the average associated vocabulary problem. of unique annotations is pretty consistent between 2 to 3 annotations. This signifies With both hypotheses supported from our that there are constant new annotations case study, we propose the user profiler of being added by the users to enhance the PIR systems to synchronize dynamically with metadata of the resources overtime. There the resource profiler. Thus, new information are bookmarks with relatively high within the associated resource profiles could maximum count of unique annotations be propagated to the user profiles to increase throughout the days which could contain the precision of the user profiler. While past valuable information. PIR systems have successfully integrated folksonomy into user profiling, none of them This would support our hypothesis that the synchronize the updates between the user information on folksonomy resources is and the resource profile. Thus, valuable enhanced overtime and thus should be taken information on resources might not be into consideration during user profiling and included into the user profile especially for the update of user profile post user early user interaction. This would be the interaction. This is particularly evident from direction of our next research to determine the maximum number of unique annotations the actual performance gain of our proposal for bookmarks in every day as there could with the addition of temporal elements.

9 Journal of Internet Social Networking & Virtual Communities

Distribution of the Delicious Bookmarks according to their Date Count

Number of Bookmarks

٣٧٠٠ ٣٦٠٠ ٣٥٠٠ ٣٤٠٠ ٣٣٠٠ ٣٢٠٠ ٣١٠٠ ٣٠٠٠ ٢٩٠٠ ٢٨٠٠ ٢٧٠٠ ٢٦٠٠ ٢٥٠٠ ٢٤٠٠ ٢٣٠٠ ٢٢٠٠ ٢١٠٠ ٢٠٠٠ ١٩٠٠ ١٨٠٠ ١٧٠٠ ١٦٠٠ ١٥٠٠ ١٤٠٠ ١٣٠٠ Number of NumberBookmarks ١٢٠٠ ١١٠٠ ١٠٠٠ ٩٠٠ ٨٠٠ ٧٠٠ ٦٠٠ ٥٠٠ ٤٠٠ ٣٠٠ ٢٠٠ ١٠٠ ٠ ١ ٩٧ ٧٣ ٤٩ ٢٥ ٥٥٣ ٥٢٩ ٥٠٥ ٤٨١ ٤٥٧ ٤٣٣ ٤٠٩ ٣٨٥ ٣٦١ ٣٣٧ ٣١٣ ٢٨٩ ٢٦٥ ٢٤١ ٢١٧ ١٩٣ ١٦٩ ١٤٥ ١٢١ Number of Dates

Fig 2. Distribution of the Delicious Bookmarks According to Their Date Count

Journal of Internet Social Networking & Virtual Communities 10

Unique Annotations Count for Delicious Bookmarks according to Days

Unique Annotations Count

٤٠٠٠٠ ٣٠٠٠٠ ٢٠٠٠٠ ١٠٠٠٠ ٠ ١ ٩٧ ٧٣ ٤٩ ٢٥ ٥٥٣ ٥٢٩ ٥٠٥ ٤٨١ ٤٥٧ ٤٣٣ ٤٠٩ ٣٨٥ ٣٦١ ٣٣٧ ٣١٣ ٢٨٩ ٢٦٥ ٢٤١ ٢١٧ ١٩٣ ١٦٩ ١٤٥ ١٢١ Day

Number of Uniqueof NumberAnnotations

Fig 3. Unique Annotations Count for Delicious Bookmarks According to Days

Maximum Number of Unique Annotations for Delicious Bookmarks

Maximum Number of Unique Annotations

٢٠٠ ١٥٠ ١٠٠ ٥٠ ٠ ١ ٩٧ ٧٣ ٤٩ ٢٥ ٥٥٣ ٥٢٩ ٥٠٥ ٤٨١ ٤٥٧ ٤٣٣ ٤٠٩ ٣٨٥ ٣٦١ ٣٣٧ ٣١٣ ٢٨٩ ٢٦٥ ٢٤١ ٢١٧ ١٩٣ ١٦٩ ١٤٥ ١٢١ Day

Number of NumberUnique Annotations Fig 4. Maximum Number of Unique Annotations for Delicious Bookmarks

11 Journal of Internet Social Networking & Virtual Communities

Average Unique Annotations for Delicious Bookmarks

Average Unique Annotations

٨ ٦ ٤ ٢ ٠ ١ ٩٣ ٧٠ ٤٧ ٢٤ ٥٥٣ ٥٣٠ ٥٠٧ ٤٨٤ ٤٦١ ٤٣٨ ٤١٥ ٣٩٢ ٣٦٩ ٣٤٦ ٣٢٣ ٣٠٠ ٢٧٧ ٢٥٤ ٢٣١ ٢٠٨ ١٨٥ ١٦٢ ١٣٩ ١١٦ Day

Number of NumberUnique Annotations Fig 5. Average Unique Annotations for Delicious Bookmarks

Potential, Impact and Application systems which would only benefit contributors from the user annotations, the Without a doubt that the step into addition of resource annotations would also personalized information retrieval (PIR) benefit users which interacted with the systems from generic information (GIR) resources through various platforms and not systems is able to meet the information need only folksonomy systems. The system’s of the users better. It is seen from our confidence and accuracy in the discussion above that most of the users understanding of users would further prefer PIR systems over GIR systems and the increase as the information about the results from the comparison of various resources and their metadata are enhanced approaches confirm it. The biggest dilemma overtime by the users of folksonomy systems. concerning PIR systems would be on the user profiling – can the PIR system understands Application of Improved User Profiling the users and their information need correctly? What are the suitable sources of The improved user proiling through web 2.0 user information to learn about the users? systems would benefit information systems The high adoption rate of web 2.0 systems (IS) in many ways and not just PIR systems. proved to be a valuable source for user Within the context of PIR systems, the information which would improve the systems would be able to meet the performance of PIR systems in many ways. information need of the users given a user query. From the query of the users, the PIR Web 2.0 systems do not only act as an system would be able to establish the context accurate source of user information but also of the query and overcome the vocabulary allow user profilers to model the multi- problem such as ambiguity by cross- interest of users, establishing the interest referencing it with information within the context (by overcoming the vocabulary use r profile. This is held true for any problem) and accounting the temporal personalized approach such as re-ranking, changes in user interest. Unlike past PIR query enhancements and recommendations.

Journal of Internet Social Networking & Virtual Communities 12

Within IS, the systems could make use of the performance of user profilers could be user profile in many ways. One of such further enhanced especially for users who application is the ability to segment and interacted with the early stages of cluster the users into interest groups using folksonomy resources before additional information from the user profile. This would information are being added. help information systems in identifying suitable user/ interest groups for effective The future direction of our research is still and efficient data mining or information focused on the user profiling aspect of PIR delivery (especially for advertising). systems especially a further look into the information gaps from the temporal By having the user profile as well, elements of web 2.0 systems. We believe that information systems could provide the users the performance of user profilers could be with suitable recommendations of enhanced by addressing this. We also look at information and resources without the need other aspects of web 2.0 systems in for users to state them explicitly. Coupled improving user profiles. One of such aspects with other user-based information such as would be the propagation of user the location of the user via Global Positioning information via social networks such as System (GPS), weather and so forth; Facebook and Twitter. We would then information systems could be truly proceed with our own PIR systems with web personalized to the users in their daily life. 2.0 elements; benchmarking it against the current generation GIR and PIR systems. Conclusion References The addition of web 2.0 systems without a doubt improve the performance of user Amitay, E., Carmel, D., Har'el, N., Ofek- profilers within personalized information Koifman, S., Soffer, A., Yogev, S. & Golbandi, N. retrieval (PIR) systems which is already (2009). "Social Search and Discovery Using outperforming the current generation of a Unified Approach," Proceedings of the 20th generic information retrieval systems (GIR). ACM conference on Hypertext and hypermedia. The main concerns with PIR systems are the Torino, Italy: ACM. system’s ability to profile the users which would directly determine the performance of Au, Y. C. M., Gibbins, N. & Shadbolt, N. (2008). PIR systems. The improvement in user "A Study of User Profile Generation from profile would allow PIR systems as well as Folksonomie," Social web and Knowledge other information systems (IS) to provide Management personalized services to a wide range of users and improving the users’ satisfaction Bischoff, K., Firan, C. S., Nejdl, W. & Paiu, R. through such services. (2008). " Can all Tags be Used for Search?," Pr oceeding of the 17th ACM conference on As with our discussion above, web 2.0 Information and Knowledge Management," systems could very well enhance the user Napa Valley, California, USA: ACM . profiler of PIR systems in terms of accuracy, adapting to interest changes as well as Cai, Y., Li, Q., Xie, H. & Yu, L. (2010). modeling the user profile. Past studies have "Personalized Resource Search by Tag-Based shown the information potential of User Profile and Resource Profile ," In: Chen, folksonomy in user profiling. As seen from L., Triantafillou, P. & Suel, T. (eds.) Web our case study of Delicious, the temporal Information Systems Engineering – WISE elements associated with folksonomy 2010. Springer Berlin / Heidelberg. systems create information gaps. This has not been explored in past research within Daoud, M., Bou ghanem, M. & Tamine- this area. When such gaps are addressed, the Lechani, L. (2009)."Detecting Session 13 Journal of Internet Social Networking & Virtual Communities

Boundaries to Personalize Search Using a Contents for Personalized Conceptual User Context ," Lecture Notes in Recommendation," LNCS (LNAI) , 4755, 116– Electrical Engineering in Advances in 126. Electrical Engineering and Computational Science , 39, 471-482. Lee, U., Liu, Z. & Cho, J. (2005). "Automatic Identification of User Goals in Web Search," Diederich, J. & Iofciu, T. (2006). "Finding International World Wide Web Conference in Communities of Practice from User Profiles the Proceedings of the 14th international Based on Folksonomies," Proceedings of the conference on World Wide Web , 391-400. 1st international workshop on building technology enhanced learning solutions for Liang, H., Xu, Y., Li, Y. & Nayak, R. (2008). communities of practice (TEL-CoPs’06). "Collaborative Filtering Recommender Systems Using Tag Information," Forsati, R., Meybodi, M. R. & Neiat, A. G. IEEE/WIC/ACM international conference on (2009)."Web Page Personalization Based on web intelligence and intelligent agent Weighted Association Rules," icect in the technology , 59–62. 2009 International Conference on Electronic Computer Technology , 130-135. Liu, F., Yu, C. & Meng, W. (2004). "Personalized Web Search For Improving Golder, A. & Huberman, A. (2005). 'The Retrieval Effectiveness," IEEE Transactions Structure of Collaborative Tagging on Knowledge and Data Engineering , 16, 28- Systems,' HPL Technical Report. 40.

Han, X., Shen, Z., Miao, C. & Luo, X. (2010). Marlow, C., Naaman, M., Boyd, D. & Davis, M. "Folksonomy-Based Ontological User Interest (2006). "HT06, Tagging Paper, Taxonomy, Profile Modeling and Its Application in Flickr, Academic Article, to Read," Personalized Search," In: AN, A., LINGRAS, P., Proceedings of the seventeenth conference on PETTY, S. & HUANG, R. (eds.) Active Media Hypertext and hypermedia . Odense, Denmark: Technology. Springer Berlin / Heidelberg. ACM.

Haveliwala, T. H. (2002). "Topic-Sensitive Michlmayr, E. & Cayzer, S. (2007). "Learning PageRank," International World Wide Web User Profiles from Tagging Data and Conference in the Proceedings of the 11th Leveraging them for Personal(ized) international conference on World Wide Web , Information Access," Workshop on Tagging 517-526. and Metadata for Social Information Organization , 16th Internl. WWW Conf. Heymann, P., Koutrika, G. & Garcia-Molina, H. (2008)."Can Social Bookmarking Improve Michlmayr, E., Cayzer, S. & Shabajee, P. Web Search?," Proceedings of the (2007). "Add-A-Tag: Learning Adaptive User International Conference on Web Search and Profiles from Bookmark Collections," Web Data Mining, Palo Alto, California , USA: Proceedings of the 1st international ACM. conference on weblogs and social media (ICWSM’06 ). Hotho, A., Jäschkes, R., Schmitz, C. & Stumme, G. (2006)."BibSonomy: A Social Bookmark Mika, P. (2005). "Ontologies are Us: A Unified and Publication Sharing System," Proceedings Model of Social Networks and Semantics," of first conceptual structures tool Proceedings of the 4th international semantic interoperability workshop , 87–102. web conference, ISWC 2005 , 522–536.

Kim, H.-N., Ha, I., Jung, J.-G. & Jo, G.-S. (2007). Milicevic, A. K., Nanopoulos, A. & Ivanovic, M. "User Preference Modeling from Positive (2010). "Social Tagging in Recommender Journal of Internet Social Networking & Virtual Communities 14

Systems: A Survey of the State-of-the-art and Conference on Artificial Intelligence and Possible Extensions ," Artificial Intelligence Applications . Review , 33, 187-209. Szomszor, M., Cattuto, C., Alani, H., O’Hara, K., Mobasher, B., Cooley, R. & Srivastava, J. Baldassarri, A., Loreto, V. & Servedio, V. (2000). "Automatic Personalization Based on (2007). "Folksonomies, the Semantic Web, Web Usage Mining," Communications of the and Movie Recommendation," Proceedings of ACM , 42, 142-151. the 4th European semantic web conference, bridging the gap between Semantic web and Mulvenna, M. D., Anand, S. S. & Büchner, A. G. web 2.0 , 71–84. (2000). "Personalization on the Net Using Web Mining: Introduction," Commun. ACM , Wal, T. V. (2005). 'Folksonomy Deinition and 43, 122-125. Wikipedia,'

Noll, M. G. & Meinel, C. (2007). "Web Search Wang, J. & Davison, B. D. Personalization via Social Bookmarking and (2008)."Explorations in Tag Suggestion and Tagging ," Proceedings of 6th international Query Expansion," Proceeding of the 2008 semantic web conference (ISWC) and 2nd ACM workshop on Search in social media. Asian semantic web conference (ASWC) , 4825 Napa Valley, California, USA: ACM. of LNCS, 367–380. Xu, S., Bao, S., Fei, B., Su, Z. & Yu, Y. (2008). Sieg, A., Mobasher, B., Burke, R., Lytinen, S. & "Exploring Folksonomy for Personalized Prabu, G. (2004)." Using Concept Hierarchies Search," Proceedings of the 31st annual to Enhance User Queries in Web-based international ACM SIGIR conference on Information Retrieval," IASTED International Research and development in information retrieval . Singapore, Singapore: ACM.