LANSPAN lectures 2018

Recognizing speech pathology from prosodic cues

Matt Coler (RUG/Campus Fryslân)

Prosodic cues can indicate the presence of neurodegenerative diseases like Parkinson’s Disease (PD). While tremor is a better-known indicator of PD, prosodic abnormalities, known as dysprosody, impact about 70% of PD patients and are among the first symptoms to manifest.

Accordingly, dysprosody holds a potentially crucial role for early detection and treatment. While expert clinicians can distinguish PD dysprosody from healthy speech, we do not know how. What in the signal is a cue to recognition?

To address this, we plan to conduct interviews and experiments with clinicians and develop a protocol to formalize implicit expert knowledge to discover how experts recognize PD dysprosody. Crucially, we begin with a semantic rather than a signal-based account of prosodic cues to identify what in the speech signal are cues to expert recognition.

Speech processing in challenging listening conditions

Susanne Brouwer (RU Nijmegen)

At first sight, listening to speech seems like an easy task. However, in everyday life it is dauntingly complex, as speech typically varies considerably across instantiations, speakers, and contexts. Such so-called challenging listening conditions can originate from the speaker. For example, listeners can encounter speakers who reduce a lot or who speak with a regional accent. They can also originate from the environment: speech can be degraded by traffic noise or babble from other speakers in the background. Finally, they can arise from listener limitations. Non-native listeners, for instance, often experience the detrimental effect of their imperfect second language during listening. Moreover, the adverse effect of such an incomplete language is exacerbated when combined with speaker or environmental degradation. In this talk, I will consider the influence of these challenging conditions on listening.

The first line of research is concerned with the question of how listeners process reduced speech such as "yesay" pronounced as "yesterday". Using a visual world eye-tracking paradigm, I will demonstrate how reduced speech can affect the lexical competition process. The second line of research examines how native and bidialectal listeners segregate speech from noise. Results on a speech-in-speech recognition task show how bidialectal status is of influence on performance. Both of these lines of research reveal how challenging listening conditions can adversely affect communication. The third line of research investigates whether such conditions also affect moral decision making, a serious task in life which could have far-reaching consequences. Preliminary results reveal differences between native and non-native listeners’ moral decisions. In conclusion, these three different lines of research show how challenging listening conditions can affect communication and possibly also moral decision making.