Summary:
What if, tomorrow, you were suddenly unable to communicate? What would you wish for? Currently, writing or a computer can replace your speech or voice, but perhaps you would dream of having your ability to communicate restored? That is the goal of the OLINPIC project: to restore communication by automatically modifying impaired speech units through the combined expertise of the humanities, computer science, and health sciences. Communication difficulties can affect patients with speech disorders or second-language learners whose articulation problems impair intelligibility. Impaired intelligibility inevitably impacts quality of life, as communication difficulties lead to social isolation and low self-esteem. Communication is an essential element of social bonding and involves extremely complex mechanisms of speech production and perception, in which we believe speech variability plays a crucial role. In its adaptive function, variability ensures intelligibility regardless of the conditions of speech production, transmission, and perception. Nevertheless, despite speakers’ compensatory strategies and listeners’ reconstruction skills, communication sometimes fails due to a lack of intelligibility and/or comprehensibility. Speech variability, at both the segmental and prosodic levels, results from numerous factors, including linguistic factors and those related to the speaker’s profile (pathological vs. L2 learner vs. control). In this context, the main challenge of our project is to distinguish between units that vary due to the speaker’s profile and those whose variability is linguistically determined (prosodic constraints, coarticulation, speech styles) in order to achieve an optimal and linguistically motivated speech reconstruction.