Voice-controlled interaction offers great potential for providing barrier-free and low-threshold access to e-government services. The short study on voice-control for e-government services in Germany, funded by the National E-Government Competence Centre (NEGZ), looks at which steps need to be considered for the implementation.
Rather than using a mouse or keyboard, the public sector can benefit from using voice control to interact with citizens, whether they are applying for a resident parking permit or filing a tax return. It is already possible to implement simple verbal status requests today. However, the key foundations for more complex interactions, such as voice dialogues that guide citizens step by step through a service provided by the authorities or through an application, are unfortunately not yet in place in Germany. What is required is an infrastructure that enables government services to be provided via voice control, whilst taking into account the exacting requirements in terms of data privacy.
We are convinced that voice control will become a popular means of interaction going forward.
“We are convinced that voice control will become a popular means of interaction going forward. So it’s all the more important to lay the right foundations today for this mode of interaction to be used in an official context”, says Dr Stefan Schaffer from the German Research Centre for Artificial Intelligence (DFKI). Schaffer is co-author of the short study and conducts research in the field of cognitive assistance systems.
The study identifies several promising scenarios for voice interaction in official processes. The focus here is on potential applications relating to tax returns. “A tax assistance system could help to capture documents or to record work trips, for example. Dialogue systems could also be used to help complete entire applications”, explains Roland Krebs, co-author of the short study and overall head of the ELSTER electronic tax return project at the Bavarian State Office for Taxation.
It’s not yet possible to use artificial intelligence on its own to implement the dialogues.
“There is huge potential in voice control, particularly for applications that all citizens use. Simply put, there is no more natural form of interaction”, adds Janos Standt, co-author of the short study and Acting Departmental Manager, Public Sector, for mgm technology partners GmbH. “It’s not yet possible to use artificial intelligence on its own to implement the dialogues. The addition of domain-specific models appears to be absolutely essential. Furthermore, clarification is needed as to where the data for the dialogue logic comes from and how it can be kept under the control of the authorities.”
The technical effort required to implement convincing voice dialogue systems is significantly higher than that for classic user interfaces. A crucial step is analysing the intention behind a user input. State-of-the-art processes use machine learning approaches to do this. Collections of up to hundreds of thousands of analysed thematic dialogues are required to teach the systems. Ideally, interactions from ongoing operations will also be continually assessed and stored in the databases. As yet, however, there is hardly any of this kind of dialogue data available for the e-government sector. Complex regulations and low case numbers also represent barriers in acquiring the requisite data.
Another challenge is operating a technical solution that meets the exacting confidentiality requirements when handling data. The use of commercial cloud-based services seems to be highly problematic due to the processing of confidential information in e-government services. It is not acceptable for sensitive information, like that involved in tax returns for example, to be passed via third-party servers. The authors of the short study advocate the use of ‘on-premises solutions’ whereby the dialogue logic, including voice recognition, remains under the control of the operator – even if this increases the cost and technical work required.
The authors recommend a step-by-step approach for integrating voice control into e-government services. This would mean that simple services such as status requests and deadline extensions could be implemented using voice control in the short term. In the medium term, it is recommended that an interview mode be implemented to prepare for larger requests. Integration with the ELSTER tax portal would also be conceivable. The long-term objective should be to enable complex applications, such as tax returns, to be completed in full and based on the context in the form of a dialogue.
Many of the scenarios described in the short study can be transposed for other services in the catalogue of measures to be implemented, as defined by the federal and state IT Planning Council in accordance with the law for improving online access to administrative services (OZG).
In summary, there is still a long road ahead until there is widespread implementation of voice control in providing e-government services. However, development should begin as soon as possible in order to maintain momentum.