This paper presents Ozone Browser 1 (OB), a JavaScript tool that uses the semantics embedded in Web documents for improving the user interaction experience.

Recently, the introduction of Microformats 2 and then RDFa 3 created a bridge between the Semantic Web and the Syntactic Web by embedding semantics into Web documents. This bridge facilitates the consumption of semantic data by third party applications such as screen scrappers, or search engines. However, until now the human-readable presentation of these semantic meta-information has been set aside. From the reader's point of view, the actual interpretation of a rendered document relies on: 1) The contextual knowledge provided by the hyperlinks contained in a document; 2) The background knowledge of the reader that helps him to bind unlinked information to known concepts. However, when following hyperlinks the reader may be disturbed on its cognitive process as it loses the information given by the previous document. Moreover, the association process with the user knowledge rather than the publisher knowledge may conduct to misinterpretation or partial understanding of the document. As computers, a reader should use the document's internal and external semantics for leading him to a better understanding of it. For making it possible, it is required to augment the document's representation by making explicit this underlying and external knowledge in a human readable way. OB is a graphical overlay that uses RDFa enabled documents for fulfilling these requirements. It uses a progressive enhancement strategy based on RDFa markups for adding different presentation layers to an existing document.

First, OB extracts the semantic content from a document by creating a lightweight client-side triple store from which inferences from the document's RDFa statements can be derived. Then, it binds a graphical overlay on the top of the current web document, highlighting all the RDFa statements containing the property tag in the document via icons. The reader can request more information about these statements by pointing on them. When doing so, inferences from inside and outside the page are executed for that statement, a contextualized view of the obtained knowledge is presented and the related information within the document (statements with the same predicate) is visualised as lines superimposed on the document itself. In order to provide these visualisations, OB dynamically generates views accordingly to the name space of the RDFa statement using a plug-in approach (Fig. 1).