Using tagflake for condensing navigable tag hierarchies from tag clouds

Abstract

We present the tagFlake system, which supports semantically informed navigation within a tag cloud. tagFlake relies on TMine for organizing tags extracted from textual content in hierarchical organizations, suitable for navigation, visualization, classification, and tracking. TMine extracts the most significant tag/terms from text documents and maps them onto a hierarchy in such a way that descendant terms are contextually dependent on their ancestors within the given corpus of documents. This provides tagFlake with a mechanism for enabling navigation within the tag space and for classification of the text documents based on the contextual structure captured by the created hierarchy. tagFlake is language neutral, since it does not rely on any natural language processing technique and is unsupervised. Categories and Subject Descriptors General TermsAlgorithms, Human Factors MOTIVATIONWith the quick growth of content over the web (e.g., the blogosphere), tag-based searches and tag cloud (sets of tags) based visualizations have become popular. Tags, whether provided by the user or extracted from the textual content, annotate online documents (such as blogs and news articles) with popular terms, thus providing an easy way to search and index them.Most visualizations of tag clouds vary the sizes of the fonts to differentiate most important tags from those that are less important (Figure 1). [2] aims creating visually pleasant tag clouds, by presenting tags in the form of seemingly random collections of circles with varying sizes: the size of the circle denotes its frequency. While quickly highlighting the most dominant terms in the corpus, these representations fall short in describing the context in which these terms occur in the collection. * This work is partially supported by NSF Award #0735014,"MAISON: Middleware for Accessible Information Spaces on NSDL". The need for contextually informed navigation within the blogosphere has been highlighted in the literature. For example, [16] observes that for large blog archives, a simple chronological order is not sufficient and a table of contents (TOC) like navigational hierarchy, depicting the topics development within the blog archive and describing how these topics relate to each other would be more effective.

Constructing topical hierarchies in heterogeneous information networks

“…The learning techniques can be broadly categorized as statistics-based or linguistic-based. Many studies are devoted to mining subsumption ('is-a') relationships [5], either by using lexico-syntactic patterns (e.g., 'x is a y') [6], [7] or statistics-based approaches [8], [9]. Chuang and Chien [10] and Liu et al [11] generate taxonomies of given keyword phrases by supplementing hierarchical clustering techniques with knowledge bases and search engine results.…”

Mining latent entity structures from massive unstructured and interconnected data

“…In the Web search domain, Yin and Shah [99] study the problem of building taxonomies of search intents for entity queries based on inferring the "belonging" relationships between them with unsupervised approaches. There is other related work for building taxonomies from Web tags with similar methodology [26]. NLP Researchers have also studied entity hyponymy (or is-a) relation from web documents.…”

Learning Hierarchical Relationships among Partially Ordered Objects with Heterogeneous Attributes and Links

“…In the Web search domain, Yin and Shah [33] studies the problem of building taxonomies of search intents for entity queries based on inferring the "belonging" relationships between them with unsupervised approaches. There is other related work for building taxonomies from Web tags with similar methodology [8]. NLP Researchers have also studied entity hyponymy (or is-a) relation from web documents, among whom Zhang et al's unsupervised approach [34] claimed the state-ofthe-art performance, though the tree structures along the relation are not exploited.…”