Looking back on AI events in March 2018

Looking back on AI events in March 2018

ESSV 2018, Ulm
Conference for Electronic Speech Signal Processing

From March 7th to 9th 2018 the ESSV, Conference for Electronic Speech Signal Processing, took place in Ulm for the first time. The purpose of the event was to bring together experts in language technology research and application. It is no coincidence that Ulm was chosen as venue for this year's conference. The first speech output and speech recognition systems were developed here by AEG-Telefunken as early as the 1970s.

Dr Felix Burkhardt, expert for text, speech and language technologies, represented Deutsche Telekom at this event. His intention was to expand his network, have an exchange about latest developements and speech research results at German universities.

The three-day event was opened with a greeting by Ulm’s University President, Prof. Dr-Ing. Michael Weber. He then handed over to Geert-Jan Kruijff, Technical Product Manager at Nuance Communications. Nuance Communications develops solutions for in-car voice control, from acoustic pre-processing to voice dialog. In his lecture, Kruijff presented basic technologies and skills needed to bring a mobile cognitive assistant to life.

The ensuing contributions, given by representatives of Daimler AG, also dealt with language assistance. Daimler AG researches and develops its own voice control systems - from dialog design to vehicle testing. For example, they presented the results of an explorative internet study on multi-intents in speech dialog systems. The system used graphical instructions to collect spontaneous user speech input. All tasks were related to automated driving, requiring test persons to formulate a varying number of intentions in their statement. An analysis of these statements revealed indications that not only dialog entry can be relevant as context for multi-intention statements. [1]

On the second day, speakers from various universities presented their research results. Dr Felix Burkhardt was particularly impressed by Stefan Ultes' contribution. In his presentation „Towards Natural Spoken Interaction with Artificial Intelligent Systems“, the University of Cambridge’s researcher gave a nice overview on how neural networks can be used in speech technology.

On the third and last day of the event, Dr Felix Burkhardt took the floor himself. His lecture „Ironic Speech – Evaluating acoustic correlates by means of speech synthesis“ dealt with the approach to evaluate ironic speech acoustic correlates by means of speech synthesis. Using a web-based listening experiment, four different characteristics were examined, pointing to ironically meant utterances in the German language: exaggerated articulation, overemphasis of the main syllable, raising the pitch at the end of a sentence and speaking with extremely low excitement. The experiment’s result: all four prosodic characteristics signaling irony led to a preference over the standard version. Consequently, recognition and ironic speech generation contributes to natural and simple human-machine communication.

In the early afternoon the conference was concluded with a closing speech by the hosts of the next ESSV. For Deutsche Telekom, Dr Felix Burkhardt took the following insight with him: " The research technology we are working with in project eLIZA is state of the art. Above all this concerns rule-based data processing support by machine-learned methods, in particular artificial neural networks".

-----------------------------------------------------------------------------------------

Reference:

[1] Daimler AG: Explorative Untersuchung von Multi-Intents in Sprachdialogsystemen
(visited 2018-05-03)

Kommentare
Add a comment
Sorry

your browser is not up to date
to enjoy this website you will need to install a modern browser.
we recommend to update your browser and to install the latest version.

iOS users, please male sure you're running at least iOS 9.

Mozilla Firefox Google Chrome Microsoft Edge Internet Explorer