Skip to Main content Skip to Navigation

Mesure de l'altération de la communication par analyses automatiques de la parole spontanée après traitement d'un cancer oral ou oropharyngé

Abstract : Speech disorders are a frequent problem after treatment of oral or oropharyngeal cancer. However, few studies focus on the consequences of this disorder on the communication abilities of patients or their quality of life. However, in clinical care, the optimization of communication abilities is a major therapeutic objective in the follow-up. In current practice, the evaluation of speech disorders gives scores that poorly predict the impact of speech disorders on communication. Automatic speech analysis, which is less variable, is a growing research field. In this thesis, we aimed to measure communication impairment using automatic analyses of spontaneous speech. We studied three aspects: the measurement of communication impairment, the automatic analysis of spontaneous speech, and the prediction of communication impairment by automatic parameters. We built a new speech corpus from 25 subjects treated for oral or oropharyngeal cancer. It includes a spontaneous speech task recorded during a semi-directed interview, but also self-questionnaires allowing the measurement of communication impairment and factors associated with speech and communication. Regarding the first aspect, a reference score measuring communication in a holistic way has been constructed. It allows to fill the lack of tools available in ENT oncology for this measurement. The second aspect concerns the automatic analysis of speech. A systematic review of the literature led us to focus on tools applicable to the analysis of spontaneous speech, which is the production context closest to everyday communication. One hundred fourty-nine Automatic parameters from the different levels of Caron's psycholinguistic communication model were extracted. Then, a selection process led to retain 75 relevant and non-redundant parameters. Finally, for the third aspect, we conducted predictive modeling of communication impairment using the selected automatic parameters (correlation of 0.83 between predicted and actual score). The correlation even reached 0.89 when including associated factors (constitution of social circles, anxiety-depression state, associated deficits, self-perception of the handicap linked to the speech disorder) in the modelling. The use of automatic speech analysis thus allows a reliable prediction of the communication impairment felt by the patients. This study opens new perspectives for the use and optimization of automatic speech recognition systems in clinical evaluation on the one hand, and the consideration of functional and psychosocial needs expressed by patients on the other hand.
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Friday, February 4, 2022 - 1:09:08 PM
Last modification on : Monday, July 4, 2022 - 8:42:56 AM
Long-term archiving on: : Thursday, May 5, 2022 - 6:59:46 PM


Version validated by the jury (STAR)


  • HAL Id : tel-03557511, version 1


Mathieu Balaguer. Mesure de l'altération de la communication par analyses automatiques de la parole spontanée après traitement d'un cancer oral ou oropharyngé. Sciences de l'information et de la communication. Université Paul Sabatier - Toulouse III, 2021. Français. ⟨NNT : 2021TOU30109⟩. ⟨tel-03557511⟩



Record views


Files downloads