English Español Français Deutsch
← Back to the list
Speech and multimodal Methodologists Directly upstream of ai100's methodology

Björn Schuller

What language technology has to evaluate when the input is not text but speech — including the emotional, paralinguistic, and individual-speaker signals that text-only methods discard.

Schuller has organized the INTERSPEECH Computational Paralinguistics Challenge for over a decade — annual benchmarks on emotion, sincerity, native-language, depression, and other paralinguistic categories that voice-based systems either pick up on or miss. His openSMILE feature extractor is the toolkit much of academic affective-speech research still runs on, and his published record covers the evaluation methodology side of that subfield from its modern beginnings. For ai100, as AI engines move from text-only interfaces into voice-mode products, this is the literature that already worked out which paralinguistic signals matter, which can be reliably measured, and which still resist automated evaluation.

Worth following when
you need to evaluate language-model behavior in voice mode and want methodology for measuring what the speech channel carries beyond the literal words.
Topics
computational paralinguistics evaluation (emotion, affect, individual speaker characteristics); the INTERSPEECH challenges body of work; affective-speech feature extraction (openSMILE).
Key works
INTERSPEECH Computational Paralinguistics Challenge organization (annual, 2009 onward); openSMILE open-source feature extractor (ongoing); affective speech-processing publications from Imperial and TU Munich.