Semantic processing based on eye-tracking metrics

This paper proposes a framework for capturing semantics from eye tracking data during the process of text skimming/scanning by readers of electronic documents and HTML user interfaces. RDFa and HTML microformats are some of the easier ways proposed by the Semantic Web paradigm for embedding semantics in web pages. XSLT transformations or specialized parsers may easily convert such documents to RDF/XML semantic repositories. However, semantics do not usually have an absolute character. Although a variety of web 2.0 oriented ontologies and microformats have been widely adopted or even standardized (Dublin Core, FOAF, XFN etc.), in order to achieve semantic interoperability, there are scenarios in which user-relative semantics are especially important, such as in the development of customization engines (web session customization, recommender systems, targeted advertising), when a certain user must only share semantics with himself or with similar persons. Having the same web document, different readers would attach various semantics and relevance to the ideas, concepts or structural blocks of the document. Eye tracking is an emerging field with multiple applications in medicine, marketing, cognitive sciences and others, which allows the extraction of data regarding the eye activity of a user during human-computer interaction. Eye tracking data is valuable in measuring reading patterns, user experience and reflects the specific parts of an image or documents that attract the user's interest.