Transcription of court sessions is a tedious and time-consuming activity. According to Viljar Peep, the Deputy Secretary-General of the Judicial Administration Policy Department, a single criminal case with total hearings of about 14 hours of audio recording could take an average of 112 working hours to transcribe. Meanwhile, all that burden falls on the court clerks. Salme, a speech recognition tool developed in partnership with IT company CGI Estonia and language-tech company Tilde, has been introduced to Estonian courts to simplify and optimise this operation.
With 92% precision, the programme is designed to increase efficiency and accuracy in court sessions transcription. “Whether the court hearing takes place in a physical courtroom or online, Salme’s duty remains the same. This solution was introduced to help save transcription time, particularly in the case of long court hearings that last well over six hours,” says Raivo Tammus, IT Manager of the Estonian Courts.
Integrated with the X-Road
Estonia’s e-justice system, under the management of the Ministry of Justice, is made up of three interconnected components. They are the e-File Central Information System, the Court Information System, and a public portal for members of the society. Now part of the Court Information System is Salme. “The court proceedings’ data is received over the X-Road from KIS, the Court Information System. And after the hearing, the recording, transcript, and protocol are sent back to KIS, again over X-Road,” Raivo explains.
According to Mihkel Kivisild, a Software Developer at CGI Estonia, Salme is a double-threat, user-friendly and intelligent. “While a court session is ongoing, the online transcription service created by Tilde responds with the transcript of what is being said, with only a few seconds delay. It divides texts into blocks and then fixes them based on context information,” he says. Notwithstanding, development endeavours are progressing to increase the solution’s intelligence and user-friendliness by the end of February 2022.
The one-stop-shop for court proceedings
“At the beginning, the operator manually adds the attendees to the court session, querying them over X-Road, and then assigns the microphones to the respective users. By assigning the microphones, Salme knows who is talking and which channel needs to be recorded per time. And then there is the transcription side of it,” says Mihkel, describing Salme’s mode of operation. There are two transcription options: the real-time transcription option and the offline transcription service, which takes more time but is more accurate.
“There’s also the option to insert and manage notes related to the court hearings. The notes have their timestamp information so users can easily access the specific audio recording, transcript, and note of a particular time they need. Salme also includes a built-in spelling functionality to ensure the accuracy of spellings,” he expounds.
CGI Estonia: decades-long collaboration with the public sector
With consistent collaboration with the Estonian government for over 20 years and still counting, CGI Estonia has undoubtedly contributed several blocks to developing the e-Estonia ecosystem. “We have been working with the Estonian government since the 90s. We have partnered with the Estonian government in implementing diverse public sector solutions for many decades. Among others, we have worked with the Ministry of Justice through the Center of Registers and Information Systems on several occasions before collaborating on Salme,” says Erik Tiits, the company’s Director of Public Sector Consulting Services.
Aside from the ongoing collaboration on Salme, CGI Estonia is also working with the Centre of Registers and Information Systems and the Prosecutor’s Office to develop an upgraded version of the Prosecutor’s Information System. The seasoned IT company also partnered with SMIT, the Ministry of the Interior’s IT department, to create the Emergency Response Centre solution and the e-Call solution. The company is still very involved in improving and upgrading these solutions as new technologies emerge.
More e-assistants to come?
Why bother humans with tedious yet demanding tasks when machines can do them faster and better? Knowing this, the Estonian parliament was the first state office to introduce a speech recognition solution to streamline its operations. The programme is named Hans. “Although they are quite similar, Salme is customised for court proceedings,” Raivo comments. Being a state-funded software and one that can increase efficiency by a great deal, Hans and Salme-like solutions will prove helpful in other public sectors.
“There are a lot of use cases that we are already seeing amongst our current clients. Besides court hearings, this solution could be useful for police interrogations, emergency response calls, conferences, and similar public sector operations. Anywhere there is a significant amount of voice-based data being generated, this solution is worth considering,” Erik points out. “But we have to emphasise that Salme was trained based on court and justice-specific vocabulary. So if this solution is to be implemented in other sectors, it would need to be retrained with those sectors’ industry-specific vocabulary,” he notes.
100% digital competence by 2024
Raivo says one of the Ministry of Justice’s goals towards the next coming years is to achieve 100% digital competence. “At the moment, we still have too much paperwork. Meanwhile, we want to increase our digital competence. By 2024, we hope to have phased out printed documents. We want to be 100% digital – no paper in people’s hands. The court hearings will only include screens, Salme, and the court hearing transcripts, whether the court hearing is in the courtroom or remotely,” he concludes.