You are here

INFINITI (Information retrieval for information services)

Streamwatchr, monitoring Twitter to find out to which music people are listening. Streamwatchr offers real-time insights into music listening behaviour around the world (video at the end of the page)

Digital data have entered everyone’s life, all professions, each business and all relations. It is hard to imagine any such driver in the past: the knowledge of Internet, the media of the digital channels, the instant social media, the transformation of the news, the organizational power of smart phones, the globalization enabled by the network, the ease of interaction, the support of apps which is starting to come and the statistics of professional live in Big Data.

Continuous change

The information retrieval landscape is also changing rapidly. The type of content for which intelligent information retrieval methods are needed is subject to continuous change: from a strong focus on edited content to a much broader scope that includes highly dynamic and often highly subjective user generated content. Our information needs are changing alongside: serving simple informational queries is a thing of the past. We now expect the search engine to support us in our daily tasks, whether work or entertainment related, to find facts, to find experiences, all with increasing levels of personalization.

In this changing landscape INFINITI develops new algorithms for real-time semantic analysis and for analyzing subjective content mined from social media. It pushes the state-of-the-art in mobile and cognitively informed multimedia access and it creates unsupervised ways of assessing the quality of retrieval results from implicit feedback.

Biggest results so far


​We have developed the web application Streamwatchr, that monitors Twitter to find out to which music people are listening. Streamwatchr offers real-time insights into music listening behaviour around the world. Streamwatchr More.

ICT-Science question: the core scientific challenge that Streamwatchr addresses is how to interpret highly dynamic user-generated texts. Streamwatchr maps the content of text messages to a knowledge base in real-time.

Involved COMMIT/partners: 904Labs, UvA and Eagerly Internet. 

Scan and print yourself in 3D

You do not have to invest thousands of euros on 3D scanners or you do not have to spend time to learn complicated 3D CAD systems. This demo provides you a low-cost and real-time 3D scanning solution.  With our software you can create a 3D-selfie in less than a minute. More.

ICT-Science question: the main challenge is how to use a low resolution, low cost depth camera to perform real-time 3D-scanning in order to reliably print a 3D-model. The present high resolution 3D-laser scanners cost about twenty thousand euro. These scanners are not affordable by consumers and they are not user friendly. Moreover, the necessity of post-processing makes them less applicable for daily usage.

Involved COMMIT/partners: 3DUniversum and UvA.

Exploration tool for investigative journalists

Given a pile of documents, a journalist faces the task to dig deeply into the various lay­ers that could yield a newsworthy story on political and other powers. Especially for investigative journalists we have devel-oped the exploration tool FIDO (Freedom of In­formation Document Overview). FIDO supports the exploration, search and analysis of large collections of texts, such as the ones that jour­nalists can get through Freedom of Information Act-requests (WOB in Dutch).

ICT-Science question: how can we discover in a large set of documents the key entities? How can we find the evidence central to these entities and the contextualization of mentions of the same entities in other sources? Specific technical challenges are the heterogeneity of resources to be integrated, the multi-lingual character inherent to the real-world setting and the noise resulting from optical character recognition (OCR - used to automatically generate a digital text by scanning an analog, printed text). Unique about FIDO is that it tackles these issues ‘on-the-fly’, instead of via a priori analysis. More.

Involved COMMIT/partners: Taalmonsters, Radboud University, University of Twente, Erasmus University, CWI, Spinque.

Enriching use and experience of a radio

Until now radio content has been disconnected from other audiovisual and textual informa­tion. Our demo shows that radio can be much more then listening only. Radio can be made searchable to find a certain quote, person or topic. The experience of listening radio can be enhanced presenting related images from other information sources on a screen. Reversely, radio content can also be linked to information on the web. A written news story might be enriched with a related radio interview. We improve the use and experience of radio content by automatically processing audio files and combining the output with the latest audio search technology. More.

ICT science question: how can information technology add value to media by enabling concepts that increase the use of audiovisual content? The ICT-solution should look for a combination of speech technology and information extraction to decode the spoken words automatically and use these as labels for searching and linking func­tionalities.

Involved COMMIT/partners: Beeld & Geluid, NPO, VPRO, University Twente.