Tag and word clouds as means of navigation support in social systems

Tag cloud is a visual interface that summarizes an underlying data by depicting the most frequent terms (also called as tags) from the dataset. Tags are linked to documents that contain given tags selection. A majority of tag clouds consists of the most frequent tags from a corpus that are alphabetically sorted. However, it has several drawbacks: frequent tags do not have to be relevant for all users, a vast number of terms are semantically similar hence a cloud contains many redundant depictions, an alphabetical sorting of tag cloud does not allow users to discover relations between terms. The objective of this PhD project is to propose, implement and evaluate novel tags selection methods for more relevant, diverse and novel tag clouds. Enhanced relevance of tag clouds should increase the likelihood that user will accomplish a given information retrieval task. Improved diversity and novelty of tag clouds should result into coverage of the entire spectrum of topics from folksonomy resources. Another objective is to expand a set of well-known synthetic metrics (i.e, Coverage, Overlap and Relevance) with new metrics that will capture diversity and novelty of tag clouds. Next ambition is to develop methods for tags clouds generation on top of social networks such as Twitter or Facebook. The objective is to propose words selection methods that will cover as many diverse subtopics from the underlying set of documents, tweets or statuses. The motivation is to minimize the user effort to skip redundant content.