Abstract:

Stakeholders play a fundamental role in carrying out planned tasks, achieving goals and managing them effectively, therefore it is important to know and collect opinions from a wide group of them. Limiting interpersonal relations resulting from the pandemic, as well as costs of business, lead to a rapid increase in the use of information technology in the contacts of organizations with their stakeholders. Consequently, effective speech quality management is essential to ensure the smooth running of ICT-supported stakeholder engagement and consultation processes. The main purpose of this paper is to propose a model based on selected AI methods supporting quality management in stakeholder consultation through online interview based on audio recordings. This model fills the gap resulting from the lack of such solutions used during stakeholder consultation, which requires the collection of audio content of an appropriate quality from the digital space. The correct course of consultations and the effective use of their results depend on the quality of the speech data gathered. The proposed model makes it possible to support quality management and eliminate data that does not meet the requirements for the level and legibility of the recording. Audio data from individuals or fragments of illegible records may be automatically identified and deleted. In addition to theoretical considerations and the model, the results of empirical research are also presented in this paper, including a discussion on the errors in the classification of empirical speech data. The considerations are supplemented with the results of a comparative analysis of models based on two methods: Gradient Boosting and Fully Connected Deep Neural Network.