Azure AI Speech

Municipality of Waalwijk

Making an automatic record of the hearings of the Objections Committee.

Last change on 9th of April 2026, at 14:43 (CET) | Publication Standard 1.0

• Publication category: Other algorithms
• Impact assessment: DPIA
• Status: In use

General information

Theme

Organisation and business operations

Begin date

01-2026

Contact information

[e-mail]

Responsible use

Goal and impact

Converting audio clips from hearings to text makes the content accessible, searchable and suitable for analysis and reporting. This increases transparency, findability and reusability of information for stakeholders and supports accountability in decision-making.

Considerations

Important trade-offs are safeguarding privacy versus saving time, better reporting and reducing the administrative burden complemented by attention to transcription errors and correctly capturing nuance and context.

Human intervention

Human intervention consists of entering the audio recording, checking, rephrasing and business summarising the transcription, paying attention to accuracy, completeness, nuance and context. In the process, any errors are corrected and privacy-sensitive information is carefully assessed.

Risk management

Risk management includes that audio recordings are not stored in the tool and that data remain within the secure environment of the Waalwijk municipality according to the applicable retention period. The data is not used for training and is only processed with the consent of the resident. In addition, human verification takes place by reformatting and checking the transcription. We share the record with the objector for verification and possible corrections.

Impact assessment

Data Protection Impact Assessment (DPIA)

Operations

Data

During the hearing of the Objections Committee, personal data such as name and contact details are processed. Necessary for the hearing, special personal data may also be processed, such as medical or financial information. The content of the stakeholders' input, including arguments and statements, is also recorded for a complete and accurate record.

Technical design

The algorithm helps to convert a sound recording to text as accurately as possible. It does this by using speech-to-text speech recognition technology, where the system is trained on large amounts of sound data to recognise spoken language and convert it to written text. This takes into account different speakers with accents, speaking speeds and background noise.

Using an additional algorithm, this transcript can then be automatically analysed and summarised into an accurate and structured report. In the process, relevant topics, key points and context are extracted from the text to create a comprehensible and usable representation of the conversation.

External provider

Microsoft