KIT entwickelt KI-gestütztes Transkriptionssystem | ZDNet.de

Automatische Verkennungs- und Oversetzungssysteme wie der Lecture Translator de Karlsruhe Instituts für Technology (KIT) können das Prochene in Vorträgen or der Vorlesungen in Echtzeit in einen Text in meereren umwandeln. View the system of information for study results with underlying and Australian study results.

You can perform a whole range of work and learning activities during intelligent processing and archiving of spoken texts. If you wish to read more about it, use the KIT lecture translator to add new functions. Automatische Erkennung des Gesprochenen in mehreren Sprachen gleichzeitig, Textsegmentierung und Titelgenerierung in Echtzeit, Zusammenfassungen und Verlinkungen zu Fachbegriffen und Abfragen des Gehörten vereinfachen nun das Verstehen en effiziente Aufarbeiten von Vorlesungen.

Better overview in documents

“With the automatic simultaneous presentation of the lecture, translators have brought their spoken presentations closer to an international audience. This power is usually only available to 15 percent of the listening capacity. With the new AI workers who have never spoken again, there are many different barriers,” says Alexander Waibel, professor of computer science at KIT.

Often the text of spoken is often automatically translated, so that you can read whether you can read longer the text and the Zwischenüberschrifts longer – and how you can use the Vortrag or the Vorlesung mündlich vorgetragen wurde. If the nutrition of the previous time is a lot of effort, then the man who goes to the other side the previous time is waibel.

The further development of the lecture translators should be assisted here. The researchers have enabled more new automatic functions with “Smart Chaptering”, “Summarization”, “Q&A” or “Auto-Links”. If you send a new artificial intelligence (AI), the text is automatically recognized, the text in a transcription is drawn out longer and automatically identified, chapter headings contain an important core.

The device has an acoustic regeneration, with the Nutzerinnen and Nutzer one of 18 languages ​​can be selected. Add automatic links to the program if you use relevant searches in the previous description or Wikipedia, with the help of the study results previously available. “With our new AI models, conversations and lectures can be better structured and so videos can be integrated into easily navigable chapters”, so Waibel. If there is a problem, it is not possible to pay attention to the foreground.

18 languages

The work done by the Research Team in the Lecture Translator is integrated, with the KIT Vorlesungen in Echtzeit being automatically transcribed. Kapiteeleinteilung, Titelgenerierung, Absatzgestaltung, Zusammenfassungen mit Verlinkungen – auch jeweils online und offline anwendbar – erweitern nun de Service des Lecture Translators und vereinfachen die Arbeit mit dem Material.

Concrete application possibilities find the technology for Content Creator, Study Performance, Learning and Podcaster, the first audio and video content of the first structures. “Users and users can browse through videos and other navigation systems more efficiently, find relevant summaries a core content compact and efficient inheritance – they have a great opportunity to pay attention to the details as quickly as possible,” says Waibel.

The research on the project ‘How is AI changing science?’ is one of four years that the Volkswagen Foundation is supported. Project involvement came from the KIT of the University of Bonn, as well as the University of Vienna.