ArticleRead (18): The Dark Side of the Semantic Web

The Dark Side of the Semantic Web, By James Hendler, in IEEE Intelligent Systems, 2007, Vol 22, No 1, 2-3 ; and his presentation.

The short analysis on the relationship of Semantic Web and AI research fields in observable Web 2.0 is given. The "dark side" of the Semantic Web is treated from the point of AI community that they are lack of experiences on social aspects of the Web (e.g. the use of social tagging).

Hendler emphasizes that the theme of AI – “that which looks easy in the small is often much harder in the large” . On the other hand, the catch phrase in Semantic Web – “a little semantics goes a long way” is on revealing the main trend: RDF/OWL in practice.

Two benefits for building RDF triple stores include:

1. RDF enables you to store data in a flexible schema so you can store additional types of information that you might have been unware of when you originally designed the schema.
2. RDF helps you to create Web-like relationships between data, which is not easily done in a typical relational database.

In particular, Hendler notices that the main design principle for RDF is that having unique names for different terms has a great impact on the Web. However, the challenge goes to the question of two URIs whether they are indicating the “same” , the “different”, “different parts of the same thing”, or the others.

ArticleRead (6): The folksonomy tag cloud: when is it useful?

The folksonomy tag cloud: when is it useful? By James Sinclair and Michael Cardew-Hall ,Journal of Information Science 2008 34: 15-29

With the assumption of folksonomy systems affecting user perceptions and patterns, it is interesting to see what empirically can be found from a user interface point to see how tag cloud impact users in information exploration. Sinclair and Cardew-Hall in this paper clearly conclude their findings in small-scale enterprise context, which supports arguments of Mathes (2004) and Brooks & Montanex (2006), that the usability of tag cloud is a social navigation aid tool when broad, general or vague information exploration is taken up. Increasingly, evidences from empirical survey support the function of tagging for broad categorization. [e.g. Noll and Meinel, 2007]

A proper appreciation of this research with which the need to evaluate Tag Cloud in its usability is asserted in its visual summary design, and its ability to serve for non-specific information discovery. Such results are also given weights to a substantial literature reviews of many pro-and-con characteristics of tag clouds. Here, we try to summarize both ends in usability and sociability analysis in the table below.


A very interesting section of this article is that: only 2 out of 89 participants with high level computer background in their experiment are familiar with the tagging mechanism. This percentage is surprisingly low while comparing to the overview that almost one third of online American users have used tagging mechanisms. Out of most curiosity is that since this study is a research on user patterns and perception, there is a missing data analysis to undertake. While the study has concluded that the cost of a query is reduced in the tag cloud scenario (compared with more typing efforts in search box), should the mean tags tagged per article of each participant need to be considered as one of the factors in the cost analysis? Indeed, this remains a question to explore.

ArticleRead (4) : Collaborative Tagging and Semiotic Dynamics

Collaborative Tagging and Semiotic Dynamics, By C Cattuto, V Loreto, L Pietronero ,Arxiv preprint cs.CY/0605015, 2006
From the Cover: Semiotic dynamics and collaborative tagging, Proceedings of the National Academy of Sciences, 2007 - National Acad Sciences

In “Collaborative Tagging and Semiotic Dynamics”, Cattuto, Loreto and Pietronero set down what a user pattern looks like in a social tagging system through empirical statistic analysis of tag co-occurrence. The Yule-Simon model on probability and statistics basis has been used to investigate the long-term memory of users’ tag-vocabulary activities in one of the social tagging system, del.icio.us. A semiotic conceptual model for the tri-partite graph to structure a post as (user, resource,{tag}) is proposed. Therefore, the tri-partite concept which is original from semiotic dynamic literatures is illustrated in the tile as a highlight.

In order to overcome the need for complexity of experimental data, this analysis procedure employs a tag-centric construction view on del.icio.us system. By factoring out two parameters of (users, resource) and adding the set of time parameter from the post, the results of co-occurrence of tagging activities are shown to be consistent with available theoretical calculations in Power Law and Zipf’s Law. Typical applications of utilizing these two statistical theories are well recognized in phenomenon analysis such as in natural language; self-organization and human activity; access patterns; as well as memory–kernel of cognitive psychology. This joint experiment with the well-proved theories offers an alternative method to explore social tagging phenomenon, and for our review to add value on their ideas and research attentions on user behaviors and semiotic concept.

Controversially, however, the research method is likely to be criticized from the semiotic point of views.

First of all, the confusion of two semiotic schools is presented. The authors intend and develop the tri-partite graph from semiotic dynamic concept which follows Charles Sanders Peirce’s (1839-1914) sign theory remarkably in its basic triadic relation within a sign, namely (Represent, Object, Interpretant). The authors have attempted to adopt this triadic elements and rephrase them from (forms {words}, referents {objects}, meanings {categories}) of Steels and Kaplan (1999) to (user, resource,{tag}) in social tagging concept.

However, the authors’ reference of semiotic dynamic is the work of Ke et.al (2002) who adopts the Ferdinand de Saussure (1857–1913) school of semiotics which takes a sign being constructed within a dual relation (signifier, signified). Since these two semiotic schools have been in debates for decades, the authors adopting these two papers as their definition for semiotic dynamic may lead to confusion in general.

Picture 1: Steels and Kaplan (1999)'s Semiotic Dynamic

Secondly, their proposal for the tag-centric calculation method contradicts their own arguments favoring semantic context. In such context, semantic meaning is supossed to deal with the same Object (the same resource / bookmark in this research) to investigate the relation between different users and users' tags on their co-occurrence. Picture 1 shows the original method for the co-occurrence of items for their semantic meaning in Steels and Kaplan (1999) ‘s semiotic dynamic which the authors have cited from. Picture 2 shows the authors' method for calculating the co-occurrence of tags. Different objects (resources /bookmarks refering to) and different users(interpreters) are ignored in this case. The main focus is on the different tags’ relations, namely frequency and co-occurrence especially in low-high rank tags. Note that our review is not to argue that the authors’ work cannot result in user’s activity patterns since the Power Law and Zipf’s law have been well-proved in such domain. To be specific, if the authors’ work is not in the semiotic dynamic domain, the contradiction may not be this tremendous.



Picture 2: the authors' semiotic dynamic?