A Semantic Visual Text Abstract

A Semantic Visual Text Abstract

Custers, S.

Aprile, W.A. (mentor)
Van Mourik, F. (mentor)

Industrial Design Engineering

Industrial Design

Master of Science Design for Interaction

2011-06-24

Today’s readers want a personal, filtered text abstract that gives an objective insight in a book, instead of the current subjective blurb on a book backs. The abstract’s aim, is to analyze a book on a semantic level, showing more than just an analytic visualization of text. A semantic visual text abstract interprets text and returns a filtered user-specific view on a book. To analyze text the field of subjectivity analysis provides several tools. The semantic orientation of phrases can be defined by a statistical approach called SO-PMI-IR (Semantic Orientation Pointwise Mutual Information Information Retrieval). This algorithm categorized words on a (continuous) positive-negative scale. This semantic orientation analysis, is performed on the book’s most important phrases. These phrases are defined by the TF-IDF (Term Frequency Inverse Document Frequency) keywords of a book. TF-IDF is a popular tool for extracting the most important and unique words from a book. This combination of unique book information and semantic orientation provide a measure for comparing books on content. Apart from a glimpse of a book’s content, readers prefer an insight in what other readers think about the book. Reviews are a perfect platform to measure this popularity. Reviews are placed over time, have a rating and a certain usefulness towards the reader itself. This usefulness is defined by creating user profiles. Neighboring users (people with similar reading taste) are likely to produce more useful reviews, compared to users that differ a lot with a reader’s personal preference. A last measure within a semantic visual text abstract will define the amount a book fits a reader’s taste, also know as a recommendation. This is done by using these pervious described usefulness values and calculate its average. A readers personal average is therefore not based on all the data but on a selected neighboring group. This means the result is always user specific, as these groups and profiles always differ. The final design of the semantic visual text abstract, uses these three blocks (book contents, reviews & recommendation) to create a visual abstract of a book. The final visualization consists out of two layers, the visual layer and the underlying (textual) content layer. The first layer’s purpose is to make quick comparison possible whereas the second layers provides the much desired objective in-depth information. In the end, a semantic visual text abstract provides readers with an objective visual tool that offers a look in a book’s content and context, helps them compare books mutually and last but not least makes the decision process a little bit easier.

data visualization
semantics
book abstract

http://resolver.tudelft.nl/uuid:009e400d-bea1-4ffb-b408-6542c6cfbd57

Campus only

Student theses

master thesis

(c) 2011 Custers, S.