New Demo: Search in Text & Video
Find and explore content not only in text sources but also in audiovisual data. Several modes of visualising (“widgets”) let you approach topics in a new and innovative way. The first demonstrator of the EU project EUMSSI is now out for testing.
What is EUMSSI? EUMSSI – Event Understanding through Multimodal Social Stream Interpretation – is an EU-funded project developing technologies for identifying, aggregating, and structuring data, including traditional multimedia content as well as social media.
In EUMSSI we have assembled a large collection of data around one specific topic, which is fracking. Fracking is a technique to recover gas and oil from shale rock. You can find content in various sources including Deutsche Welle (articles, audio, video), The Guardian, Die Zeit, El Pais, Le Monde, Twitter and YouTube. The EUMSSI dataset, however, is not restricted to that topic, allowing for flexibility and scalability.
How to get started?
Try and look for a term like “fracking”. Click on the different tabs to view data from different angles, for example a map view, people view or on a timeline. You want to select tweets based on polarity? Use the polarity tab to get a first impression on “positive” versus “negative” tweets. Click here for more explanation on the various “widgets” (PDF).
The demonstrator deploys technologies such as:
- Video analysis: video recognition, face recognition, person recognition and identification, optical character recognition (OCR)
- Audio analysis: speech recognition, speaker recognition, speaker diarization
- Text analysis: named entity linking, sentiment analysis, keyphrase and topic extraction.
For more information on the EU funded Project visit the EUMSSI website.
DateTuesday, December 15th, 2015
Tagsaudio analysis, author: Kay Macquarrie, automated storytelling, EUMSSI, text analysis, video analysis