Dialog+: Intelligent House Control with Speech
Final Project Workshop
We cordially invite interested researchers and companies to our final project workshop n December 7, 2015 at TU Darmstadt.
Description
Dialog concepts for enabling a smart command & control interface via voice in a house environment. In combination with a semantic background information service for assisting the user during discussions. The goal for the user is that the system gives additional topic-related information, but not disturbing the user. The challenge here is find out possibilities for the design of a smart assistant system at home.
The project is funded by the BMBF in "KMU-innovativ" Technologiebereich "Mensch-Technik-Interaktion für den demografischen Wandel".
During this project we have started to build open acoustic models for German: Open Source Acoustic Models and a prototype for Ambient Search, a just-in-time document recommendation system for (spoken) conversations.
External Partners
- BMBF (project agency)
- VDI | VDE | IT (project leader)
- Tellur
- CIBEK
- SIBIS Institute
- Ehlers, Ehlers & Partner
Prototype Ambient Search
In the course of the BMBF Dialog+ project, we developed a prototype called Ambient Search. The idea of Ambient Search is to unobtrusively provide relevant additional information in conversations. An information source can be for example articles from Wikipedia, which are then recommended in real-time according to the current content of a discussion. We think of conversations in a broad sense, as projected use cases are meetings, lectures, private discussions or even TV shows.
The prototype is available for German and English, can use open source acoustic models with Kaldi or rely on additional subtitle streams if available (e.g. for TV shows).
People
- Prof. Chris Biemann
- Benjamin Milde
- Dr. Dirk Schnelle-Walka (now at S1nn GmbH & Co KG)
- Stephan Radeck-Arneth, Msc. (now at MaibornWolff GmbH)
Selected Publications
[1] Open-Source German Distant Speech Recognition: Corpus and Acoustic Model. Stephan Radeck-Arneth, Benjamin Milde, Arvid Lange, Evandro Gouvea, Stefan Radomski, Chris Biemann, Max Mühlhäuser. In: Text, Speech and Dialogue. Proceedings of the 18th International Conference TSD2015, Plzen, Czech Republic, to appear, Springer, September 2015. [2] Sprach- und Gestensteuerung für das Smart Home. Christoph Ruf, Jürgen Striebinger, Dirk Schnelle-Walka. In: JavaSPEKTRUM, März, 2015 [3] An Open Source Corpus and Recording Software for Distant Speech Recognition with the Microsoft Kinect. Dirk Schnelle-Walka, Stephan Radeck-Arneth, Chris Biemann, Stefan Radomski, In: Proceedings der 11. ITG Fachtagung Sprachkommunikation, September 2014. (pdf) [4] Towards Ambient Search. Stephan Radeck-Arneth, Chris Biemann, Dirk Schnelle-Walka. In: Proceedings of the LWA 2014 Workshops: KDML, IR, FGWM, Aachen, Germany, 8-10 September 2014, Springer, September 2014. (pdf) [5] Ambient Search: A Document Retrieval System for Speech Streams. Milde, B., Wacker, J., Radomski, S., Mühlhäuser, M. and Biemann. C. In: Proceedings of COLING 2016, Osaka, Japan (pdf) [6] Demonstrating Ambient Search: Implicit Document Retrieval for Speech Streams. Milde, B., Wacker, J., Radomski, S., Mühlhäuser, M. and Biemann. C. In: Proceedings of COLING 2016 (system demonstrations), Osaka, Japan (pdf) |