Article content
Speech recognition is definitely one of the areas where artificial intelligence is showing its power and effectiveness. And what is the last thing that journalists, secretaries, and assistants wish to take care of? Word-for-word transcripts. But whether for interviews or parliamentary reports, new AI-based applications emerge as useful support tools to let the machine do the boring part of the job and allow people to focus on more demanding and intellectually challenging tasks.
In the next year, the Estonian Parliament (Riigikogu) is set to introduce HANS – AI system that will be a valuable ally to the work of lawmakers and employees of the Riigikogu. By deploying speech recognition, it will increase the efficiency and accuracy in transcripts of the sessions.
We spoke to Ahto Saks, Administrative Director at the Chancellery of the Riigikogu, to understand what the project is about. Together, we see how innovation is set to change outdated practices and increase the transparency of a countries’ institutional system.
Ahto Saks, Administrative Director at the Chancellery of the Riigikogu
What is HANS and how is it set to streamline the work of the Estonian parliament?
Currently, we are required to publish the verbatim report within one hour from the end of the sitting. It takes a lot of human effort to do that, and the burden falls on stenographers. Stenography, or in a simpler form “typist work”, is generally a task that people are not particularly motivated to do. Young people, for example, do not want to do this job for long. So why not employ more suitable technological tools for it, instead? HANS is a new information system for transcribing the plenary sessions of the Parliament. It will replace the current obsolete software in the next year.
Accuracy, transparency, openness. Which of these elements is HANS more likely to have an impact on?
Parliaments are often the most open institutions in their country, and the same applies to us in Estonia. All the information we produce is available as open data, and this helps us achieve outstanding milestones in transparency.
Using only AI in our new software would result in a loss of accuracy for now, because speech recognition is not perfect – but neither are people. We hope to achieve at least 93-95% or more correctness for plenary sittings. The rest is still corrected by humans. To achieve such a high level of accuracy, we have trained our system with 1500 hours of recordings and transcripts. Without this knowledge base, HANS would score significantly lower in accuracy. The system is also set to recognise almost all speakers: MPs, the prime minister, ministers, the president. Also, we will start providing verbatim recordings as machine-readable open data, so other systems can use that data freely as well.
How does the new AI tool interact with the digital solutions previously implemented? Will that increase the efficient functioning of democratic institutions?
The system is tailor-made, so it will be able to interact with all the other tools already existing. This enables exchanging information with the document management system, receiving data from the electronic plenary voting system, and sending data to our online platform. I also believe that our project will soon benefit from total speech recognition, particularly due to the way the system is being trained. Moreover, this type of information is also easily available to all users, as part of the open-source software provided by TalTech.
The new system is indeed set to replace the human labour employed in stenography in the Estonian Parliament, but probably it won’t have a big impact on e-government as it stands today. However, I believe it is safe to imagine that other government agencies will soon adopt this technology too. But the main advantage is out there and clear – by all means, computers can replace humans when the job is not too interesting.
What further use cases do you foresee for AI in institutional contexts?
Today we are looking into other use cases, and where we can delegate to artificial intelligence the tasks we do not want to do ourselves. We hope to make more use of automatic translations in the near future, and we are researching into intelligent data linkage systems. In all honesty, what I noticed is that we have a massive amount of information. What more can be done with it within the assembly? For example, we could consider making this data open and available. In this way, other actors outside the Parliament can use it, and generate additional benefits that go beyond internal use. Data must be public, and its availability in machine-readable formats is essential for the deployment of artificial intelligence not only at the institutional level, but across different “satellite sectors” too.
In your view, what aspects are particularly relevant for managing human-to-machine transition effectively?
Innovation is an integral part of the way we organise tasks and resources, and it is impossible to disregard it. When implementing new technologies, AI included, it is necessary to think about why and how we are doing this. Formulating use cases is crucial to understand the way solutions that are innovative by nature can be adopted. After all, digital development is not a box to tick, we don’t innovate just for the sake of doing it.
We want to use AI because it can serve better the tasks we’re called to perform. It will support legislators and, therefore, the functioning of the Parliament. Even if they do not necessarily have to, national assemblies can be the most innovative institutions in a country. Our goal is not merely to introduce AI to the Parliament. We do so because its usage is appropriate for the goals we want to achieve in the situation at hand.
The solution was developed in partnership with Finestmedia.