Integrated mobile speech recognition: the pitfalls
Speech recognition integrated into smartphones has limits in data protection, offline use and customisation. What to consider with mobile dictation apps.
The speech recognition built into smartphones is convenient but only of limited use to organisations. It works exclusively online, sends audio data to third-party servers and only knows a general vocabulary. A solution with its own speech recognition server avoids these drawbacks.
MDM as a prerequisite for mobile apps
Organisations are increasingly introducing mobile apps. Which app is right for a department is decided in a lengthy process in which the department and the IT team often hold different views. That also applies to dictation apps that convert speech into text.
An MDM solution enables remote configuration of smartphones and, above all, protects company data from access by third parties. It is therefore the most important criterion when selecting mobile apps. Many apps in the app store are ruled out because they offer no MDM integration.
Integrated speech recognition and its limits
The speech recognition integrated into iPhone, Google or Android is free speech recognition software and allows dictation into any text window. The recognised text is available immediately, the recognition rate is satisfactory and incorrectly recognised words can be corrected afterwards. At first glance this removes the detour via a typist: dictate, send, check, return.
The limitations weigh heavily, however:
- Online only. The audio stream is sent to the internet and processed on the provider’s servers. Without a good connection, recognition does not work.
- Data protection. The recognition servers are on the internet, and access to the data sent is unclear and not secured. The data is analysed in order to improve the recognition rate. This is not compatible with the GDPR. Depending on severity, fines of up to 20 million euros are possible.
- General vocabulary. It is sufficient for letters and emails. For medical reports or legal correspondence, the software recognises only parts of the speech. Unknown words cannot be added.
Speech recognition server within the organisation
Mobile speech recognition for organisations requires a speech recognition server in the company’s own network. Different vocabularies are installed there and assigned to individual users. Unknown words and text blocks can be added for users or groups.
The dictation is sent to the server via the MDM VPN gateway, recognised there and returned to the user. If the connection is inadequate, the dictation stays on the device and is transmitted automatically once network coverage is sufficient. This solution costs more than a mobile dictation app but pays for itself through the resources saved.
Key points
- Integrated speech recognition works only online, sends data to third-party servers and breaches the GDPR.
- The general vocabulary is insufficient for specialist texts and cannot be extended.
- A speech recognition server within the organisation, connected via MDM, solves both problems and also works offline.
How mobile dictation with speech recognition inside the corporate network works is described on the page ProDictate Mobile.
This article comes from the DEVACON blog and was editorially revised for the new website.
More articles
ProPlayer: playback software for dictations on Windows and terminal servers
ProPlayer is the playback software for dictations on Windows, terminal servers and Citrix Virtual Apps. Features, compatibility and use in transcription.
Speech recognition on a local server or in the cloud: data protection, advantages and disadvantages
Speech recognition servers on the local network or in the cloud: requirements, data protection, security risks and the pros and cons of both models.
Implement speech recognition: How to Calculate Economic Costs and Maximize Savings
Direct and indirect costs of speech recognition, a sample calculation for hardware and servers, and the savings from less manual data entry and correction.