Categorising Social Tags to Improve Folksonomy-Based Recommendations

Semantic Disambiguation and Contextualisation of Social Tags

“…Despite the fact that there is not a clear correlation between tag/keyword frequencies in Delicious folksonomy and the Web, we found out significant and diverse nonlinear correlations between the semantic similarities obtained with our approach in Delicious folksonomy, and the semantic distances computed with the Google Similarity Distance with Web search results. As shown in Table 8, Jaccard and Dice similarities do have a Spearman correlation value of -0.35 7 . This result supports the motivation of exploiting our semantic contextualisation approach to enhance folksonomy-based personalised web search [21] and recommendation [6].…”

“…There are approaches that attempt to identify the actual meaning of a tag by linking it with structured knowledge bases [19,1,11,7]. These approaches rely on the availability of external knowledge resources, and so far are preliminary and have not been applied to personalisation and recommendation.…”

“…García-Silva et al [11] propose a framework that selects the meaning of a tag (within a particular context -set of annotations) among a number of candidate DBpedia [3] entries using information retrieval similarity functions. Cantador et al [7] exploits YAGO ontology [20] for mapping tags to semantic concepts, and categorising them according to their purpose (describing content, context, subjective opinions, and self-organisational issues). These approaches require the existence of structured knowledge sources, and could be combined with the strategy presented herein to enrich the generated descriptions of the different semantic meanings and contexts of a tag.…”

Process Knowledge Discovery in Social BPM

“…Tagging involves the assignment of unrestricted keywords to all kinds of content [11]. These assignments can assist to build up a knowledge repository which contains related useful information on a certain topic.…”

Cold-Start Management with Cross-Domain Collaborative Filtering and Tags

“…We chose the Wilcoxon rank-sum test over the more commonly used two-samples t-test because we can not make any assumption about the normality of the distributions of the ratings. We observe that more sophisticated techniques for filtering out irrelevant tags can be implemented; for instance, as shown in [9], one can categorize tags in different semantic groups and then treat them as having different relevance weights. Denoting with T R u (i) the set of relevant tags assigned by user u on item i, UserItemRelTags predicts user ratings as follows:…”