„Hätte für meinen Geschmack gerne etwas expliziter sein dürfen, also Hard- statt Softcore, but that’s just me.“
— René Walter, Nerdcore

Sneak Preview?

johannes p osterhoff - Sneak Preview?

The semex interface facilitates users to instantly know what to expect in faceted video searching.
Paper to be presented at IVDW during Informatik 2011.

With my esteemed colleagues Dr. Harald Sack, Jörg Waitelonis and Joscha Jäger, I recently wrote a paper on our project semex—a semantic search engine that supports exploratory video retrieval.

Read the paper as PDF, 1 MB.

semex combines entities and content-based suggestions to support semantic search. The semex interface provides facet filters, entity recommendations, a pagination and browsable search results; all of these elements are interconnected by instant “linking and brushing”. Hence semex enables instantaneous visual feedback on how a selected search facet will affect the displayed result set. This unique feature facilitates a quicker and clearer search process compared to other systems that need to perform new search queries to achieve a similar result. By making the subsequent results available at fingertips, the semex interface allows users to pursue related exploratory search strategies that lead to a genuinely responsive and quick browsing experience.


Disambiguation on user input (1), the “Context Explorer” widens the search result (2),
facets derived from the result set (3) and the results themselves (4).
Brushing (5) over a facet provides visual feedback on how the result set will be changed (6)

With the introduction of Google Instant, popular web information retrieval has become quick. Immediately after entering a keyword into the search field, suggestions are shown in a drop down box and the result set is constantly updated during text input. Thus, Google provides instantaneous feedback on keyword search and resolves what once have been successively received web pages into a continuous flow of user input and system feedback. As e.g., the video search of Bing plays videos already in thumbnail preview. In this way it transfers the users’ decision making process from the sites actually containing the video content to the page of the video result set itself.

While most keyword-based search systems are optimized to narrow down a huge data space to the most suitable results and present them in “ten blue links”, exploratory search aims at finding results, which are not considered to be related at first glance. Similar to the concept of the flaneur, an urban wanderer of the late 19th century, who leisurely walks through the more and more modernized urban landscape and draws inspiration from it, exploratory search assists the user in navigating the growing information spaces and lets her choose between alternatives, move along paths, and move back to discover sidetracks and to choose alternative ways. The growing amount of structured data on the web promises to enhance exploratory search by taking into account the actual meaning of the information and its semantics.

Even though speed in terms of system feedback and an upfront but deep gaze into video result items have become available in popular web information retrieval, in the case of exploratory semantic search, instantness and an early preview into represented items on a semantic level, is highly required but only slowly taking shape.

Exploratory semantic search is based on facets and content-based recommendations, enabling the user to better refine and broaden search queries. Extracting entities from vast data collections and to find meaningful representations on the interface is understood to be one of the most important challenges in search today. User interaction overall needs to become more intuitive and lightweight. The success of exploratory semantic search heavily depends on accessible and user-friendly interfaces.

Taking these aspects into consideration, our paper presents the semantic search engine semex, that facilitates its users to perform state-of-the-art exploratory video retrieval. The objective of the interface design was to support its users with a quick feedback on selected facets that encourages exploratory search.

We would like to thank Prof. Mario Doulis, Jörg Frohnmayer and Stephan Schröter from Merz Akademie Stuttgart for their valuable feedback and contribution to the project.

semex was developed at Hasso Plattner Institute for IT Systems Engineering in Potsdam, Germany in cooperation with Merz Akademie Stuttgart during the Mediaglobe project. Mediaglobe is part of the THESEUS research program, supported by the Federal Ministry of Economics and Technology on the basis of a decision by the German Bundestag.