Asma Ounnas Semantic Social Networks Presentation

From Semantic Portal Wiki

Presentation given at CSCI 6966 Advanced Semantic Web (Fall 2008) - Lesson 5

• Speaker: Asma Ounnas
• Title: Semantic Modelling of User Interests based on Cross-Folksonomy Analysis
• Authors: Martin Szomszor, Harith Alani, Ivan Cantador, Kieron O’Hara and Nigel Shadbolt
• Conference: ISWC 2008
• URL: http://eprints.ecs.soton.ac.uk/16551/
• Date of Presentation: 2008/09/25

Questions

ID	Question	Name	Answer
Asma Ounnas Semantic Social Networks Presentation Gregory Todd Williams 1	Step 1 of the tag filtering process (Syntactic Filtering) filters out tags that are small (with length of 1) and converts tags with special characters to their "base form" (presumably into an encoding such as ASCII). This first filtering step will ignore many tags in non-Western languages (those where a tag may be comprised of a single logograph, for example). Moreover, tags in non-Western languages will fail to verify against Wordnet (and possibly Wikipedia, where by far the most articles are in English). Is the architecture presented in this paper flexible enough to deal with these challenges when faced with tags and data sources in a variety of languages, or would substantial changes be required due to a lack of suitable lexical and semantic data sources in those languages?	Gregory Todd Williams
Asma Ounnas Semantic Social Networks Presentation James Michaelis 1	Is any indication given on what degree this system seeks user feedback in determining which tags to consolidate? Given the wide variety of online identities a user may have, it is foreseeable that they would desire some variety of restriction policy on accessing certain tag clouds.	James Michaelis
Asma Ounnas Semantic Social Networks Presentation James Michaelis 2	Does this system have any mechanisms to deal with inherent information conflicts across different folksonomies? For instance, a given user may state on one music site a positive opinion of band X, and say band X is lousy on another music site.	James Michaelis
Asma Ounnas Semantic Social Networks Presentation Jesse Weaver	Section 3.4 states: "For most users, there are many categories that appear with a weight of only 1 or 2. Since these categories are the product of a tag with a low frequency, and therefore do not necessarily correspond to something which the user is particularly interested in, they are filtered out - the final category list contains only categories with a weight above the average for that user." This seems somewhat arbitrary and brings up a particular question: how frequently must a tag (or related set of tags) be used before it is (or they are) considered an "interest" of a user? Are there better ways than just cutting out all tags that fall below average usage?	Jesse Weaver
Asma Ounnas Semantic Social Networks Presentation Joshua Shinavier 1	This paper presents a collection of techniques for mapping user-generated tags to Wikipedia categories. More generally, it presents a technique for aggregating user-generated tags from accounts in multiple services and generating a user profile. However, it's not clear what we gain from these other components. The account correlation component of the architecture is intended to capture different aspects of the user's interests, but doesn't seem essential to the technique. The generation of a FOAF file as the end result of profile generation seems a little arbitrary. Would the paper have benefited from a discussion of how the generated profile can ultimately be used to provide benefit to the user (e.g. by mapping interests back to original tags in each of the source web services and using these for recommendation)?	Joshua Shinavier
Asma Ounnas Semantic Social Networks Presentation Joshua Taylor 1	In 3.3 Step 1: Syntactic Filtering the authors write "tags tjat are too small … or too large … are discarded. … Tags containing numbers are also filtered according to a set of custom heuristics: … we consider global tag frequency and discard any unpopular tags. Finally, common stop-words … are discarded." In class we've discussed some of the problems that can arise in discarding tags. Here is appears that this applies only to tags with numbers. The authors also apply a number of filters (to address misspellings, neologisms, and domain specific terminologies. Are there any effects of this that (other than what they intend) of which we should be aware?	Joshua A. Taylor
Asma Ounnas Semantic Social Networks Presentation Xixi Luo 1	the title is semantic modelling of user interests based on cross-folsonomy, but I think the main contribution of the method represented in this paper is not related too much to cross-folksonomy. What is the advantage of applying the method to cross-folksonomy?	Xixi Luo
Asma Ounnas Semantic Social Networks Presentation Xixi Luo 2	Actually I expect to see whether the profile produced from one folksonomy is close to the profile produced from the other folksonomy of the same user in the same time period. if the author can prove this, it will further confirm the assumption that tags used by users can represents their interests, since user's interests may be consist in the same time period.	Xixi Luo
Szomszor2008semantic question 1 by lebo	The authors claim that "Tagging ... (is a) knowledge management mechanism that users find easy to use and understand." What scientific, non anecdotal, evidence do we have that can convince us of this? Do /all/ users find tagging to be easy to use and understand? Are there certain contexts in which tagging is easy to use and understand, but not others?	Tim Lebo
Szomszor2008semantic question 2 by lebo	The authors claim that their "results show that far richer interest profiles can be generated." What meaningful metric are they using to measure 'richer', such that we can know that their tag-combining approach provides 'far richer' results? One suggestion would be to offer the service described in the paper and gather metrics for how many del.icio.us and flickr users request the service, suggest the service, refer back to the results, etc.	Tim Lebo

Attendees

Tim Lebo