A kratt transcribes the calls made to 112.

Transkribus transforms historical manuscripts into machine-readable format and allows for information to be searched from historical texts.

The new machine translation solution makes it easier for non-Estonian speaking citizens to understand register data. The tool helps e-Land Register users to understand detailed information such as servitudes and mortgages. The tool is based on a neural network. Machine translation establishes connections using a large database and continuously improves itself over time by refining its vocabulary and thereby increasing the quality of translation. It is important to note that the official language of the register is Estonian, which means that the translated text is informative and has no legal meaning.

With the introduction of the new e-Business Register portal, a translation kratt was developed. Upon opening a court order, the translation kratt translates all the unique information contained therein into English, directly on the computer screen. The new self-improving functionality offers a convenient way to understand the content of the document in a single view. The translation kratt takes specific terms into consideration, thereby saving entrepreneurs the valuable time typically spent on manual translation.

The researchers at the University of Tartu Institute of Technology have developed hospital robots that help reduce the workload of healthcare workers and transport samples from the ICU to the laboratory. There are many tasks that can be automated in a hospital. For example, the robots can deliver medicines and food to patients or guide them to the right room. By delegating the task of transporting samples between intensive care units and the laboratory to a dedicated transport robot, nurses and carers can use the time previously spent on logistics to take care of patients.

Pilot project for analysing the treatment invoice database.

The Health Insurance Fund has been using machine learning since late 2019 to verify treatment invoices in order to prevent and detect potential fraud or errors.

The National Library of Estonia receives a number of enquiries and requests for information on a daily basis. These enquiries often feature repetitive patterns and recurring topics, and the amount of human resources needed to address them is unreasonable. The National Library of Estonia receives approximately 14,000 emails containing questions each year. As a result of the project, a prototype of an AI-based application was developed, which was capable of sorting incoming enquiries by topic and providing answers to first-level enquiries with recurring patterns.

As part of the project, a prototype for operational flood monitoring was developed using remote sensing radar data and water level remote sensing data to improve hydrological monitoring and marine forecasting services. Hydrological monitoring includes on-site measurements of water levels and flow rates at national hydrometry stations. When combined, the water level remote sensing prototypes enable remote sensing data, the forecast model and artificial intelligence to be used to ensure better water level information for interested parties and the public. Integrating altimetry data into hydrological monitoring would help to improve the spatial coverage of surveying on major lakes and improve the accuracy of sea level forecasting.

The project resulted in the creation of three prototypes: 1) fire risk assessment and forecast; 2) fire spread algorithm; 3) determination and assessment of damage to ecosystems and damage avoided.

The kratt listens to oral presentations and generates a word cloud of the most frequently used words.

Public online speech recognition service of the TalTech Laboratory of Language Technology that can be used to convert audio files into text. The system is fully automatic and capable of processing multiple recordings simultaneously.