Jörn Hees

Text 2.0

We discuss the idea of text responsive to reading and argue that the combination of eye tracking, text and real time interaction offers various possibilities to en- hance the reading experience. We present a number of prototypes and applications facilitating the users gaze in order to assist comprehension difficulties and show their benefit in a preliminary evaluation.

A robust realtime reading-skimming classifier

Distinguishing whether eye tracking data reflects reading or skimming already proved to be of high analytical value. But with a potentially more widespread usage of eye tracking systems at home, in the office or on the road the amount of environmental and experimental control tends to decrease. This in turn leads to an increase in eye tracking noise and inaccuracies which are difficult to address with current reading detection algorithms. In this paper we propose a method for constructing and training a classifier that is able to robustly distinguish reading from skimming patterns. It operates in real time, considering a window of saccades and computing features such as the average forward speed and angularity. The algorithm inherently deals with distorted eye tracking data and provides a robust, linear classification into the two classes read and skimmed. It facilitates reaction times of 750ms on average, is adjustable in its horizontal sensitivity and provides confidence values for its classification results; it is also straightforward to implement. Trained on a set of six users and evaluated on an independent test set of six different users it achieved a 86% classification accuracy and it outperformed two other methods.

iDocument: using ontologies for extracting and annotating information from unstructured text

Due to the huge amount of text data in the WWW, annotating unstructured text with semantic markup is a crucial topic in Semantic Web research. This work formally analyzes the incorporation of domain ontologies into information extraction tasks in iDocument. Ontologybased information extraction exploits domain ontologies with formalized and structured domain knowledge for extracting domain-relevant information from un-annotated and unstructured text. iDocument provides a pipeline architecture, an extraction template interface and the ability of exchanging domain ontologies for performing information extraction tasks. This work outlines iDocuments ontology-based architecture, the use of SPARQL queries as extraction templates and an evaluation of iDocument in an automatic document annotation scenario.

Analysis and forecasting of trending topics in online media streams

Among the vast information available on the web, social media streams capture what people currently pay attention to and how they feel about certain topics. Awareness of such trending topics plays a crucial role in multimedia systems such as trend aware recommendation and automatic vocabulary selection for video concept detection systems. Correctly utilizing trending topics requires a better understanding of their various characteristics in different social media streams. To this end, we present the first comprehensive study across three major online and social media streams, Twitter, Google, and Wikipedia, covering thousands of trending topics during an observation period of an entire year. Our results indicate that depending on ones requirements one does not necessarily have to turn to Twitter for information about current events and that some media streams strongly emphasize content of specific categories. As our second key contribution, we further present a novel approach for the challenging task of forecasting the life cycle of trending topics in the very moment they emerge. Our fully automated approach is based on a nearest neighbor forecasting technique exploiting our assumption that semantically similar topics exhibit similar behavior. We demonstrate on a large-scale dataset of Wikipedia page view statistics that forecasts by the proposed approach are about 9-48k views closer to the actual viewing statistics compared to baseline methods and achieve a mean average percentage error of 45-19% for time periods of up to 14 days.

Epiphany: adaptable RDFa generation linking the web of documents to the web of data

The appearance of Linked Open Data (LOD) was an important milestone for reaching a Web of Data. More and more RDF data sets get published to be consumed and integrated into a variety of applications. Pointing out one application, Linked Data can be used to enrich web pages with semantic annotations. This gives readers the chance to recall Semantic Webs knowledge about text passages. RDFa provides a well-defined base, as it extends HTML tags in web pages to a form that contains RDF data. Nevertheless, asking web authors to manually annotate their web pages with semantic annotations is illusive. We present Epiphany, a service that annotates Linked Data to web pages automatically by creating RDFa enhanced versions of the input HTML pages. In Epiphany, Linked Data can be any RDF dataset or mashup (e.g., DBpedia, BBC programs, etc.). Based on ontology-based information extraction and the dataset, Epiphany generates an RDF graph about a web pages content. Based on this RDF graph, RDFa annotations are generated and integrated in an RDFa enhanced version of the web page. Authors can use Epiphany to get RDFa enhanced versions of their articles that link to Linked Data models. Readers may use Epiphany to receive RDFa enhanced versions of web pages while surfing. We analysed results of Epiphany with Linked Data from BBC about music biographies and show a similar quality compared to results of Open Calais. Epiphany provides annotations from a couple of Linked Data sets.

TubeFiler: an automatic web video categorizer

While hierarchies are powerful tools for organizing content in other application areas, current web video platforms offer only limited support for a taxonomy-based browsing. To overcome this limitation, we present a framework called TubeFiler. Its two key features are an automatic multimodal categorization of videos into a genre hierarchy, and a support of additional fine-grained hierarchy levels based on unsupervised learning. We present experimental results on real-world YouTube clips with a 2-level 46-category genre hierarchy, indicating that - though the problem is clearly challenging - good category suggestions can be achieved. For example, if TubeFiler suggests 5 categories, it hits the right one (or at least its supercategory) in 91.8% of cases.

Betterrelations: using a game to rate linked data triples

While associations between concepts in our memory have different strengths, explicit strengths of links (edge weights) are missing in Linked Data. In order to build a collection of such edge weights, we created a web-game prototype that ranks triples by importance. In this paper we briefly describe the game, Linked Data preprocessing aspects, and the promising results of an evaluation of the game.

BetterRelations: collecting association strengths for linked data triples with a game

The simulation of human thinking is one of the long term goals of the Artificial Intelligence community. In recent years, the adoption of Semantic Web technologies and the ongoing sharing of Linked Data has generated one of the worlds largest knowledge bases, bringing us closer to this dream than ever. Nevertheless, while associations in the human memory have different strengths, such explicit association strengths (edge weights) are missing in Linked Data. Hence, finding good heuristics which can estimate human-like association strengths for Linked Data facts (triples) is of major interest to us. In order to evaluate existing approaches with respect to human-like association strengths, we need a collection of such explicit edge weights for Linked Data triples. In this chapter we first provide an overview of existing approaches to rate Linked Data triples which could be valuable candidates for good heuristics. We then present a web-game prototype which can help with the collection of a ground truth of edge weights for triples. We explain the games concept, summarize Linked Data related implementation aspects, and include a detailed evaluation of the game.

Collecting Links between Entities Ranked by Human Association Strengths

In recent years, the ongoing adoption of Semantic Web technologies has lead to a large amount of Linked Data that has been generated. While in the early days of the Semantic Web we were fighting data scarcity, nowadays we suffer from an overflow of information. In many situations we want to restrict the amount of facts which is shown to an end-user or passed on to another system to just the most important ones.

Simplified SPARQL REST API - CRUD on JSON Object Graphs via URI Paths.

Within the Semantic Web community, SPARQL is one of the predominant languages to query and update RDF knowledge. However, the complexity of SPARQL, the underlying graph structure and various encodings are common sources of confusion for Semantic Web novices.