Speeding up judicial proceedings with Transkit, an AI-based tool that transcribes trials

The Responsible Organisation 

The Basque Government Informatic Society (EJIE) is the technological management body of the Basque Government, that facilitates the digitisation of public services, and guarantees the quality, security and continuity of the information and communication technologies (ICT services) that support them. 

The problem 

The transcription of legal proceeding is required by law and this activity places a significant burden on clerks. In fact, video recordings of trials had to be carefully examined and transcribed manually by clerks, a time-consuming and error-prone process. In addition, the manual transcription of recordings did not make it possible to provide an accurate matching between the video and the text (e.g., digital subtitles), something that could be useful for judges reviewing past trials. Accordingly, retrieving specific words, phrases, or references (especially names) within the video clip – and the text – was difficult and time-consuming, hindering the review of previous trails.  

The solution and its implementation 

The solution provided by the Basque Government Informatic Society (EJIE) is an AI tool which uses neural networks to convert voice data into searchable and structured text, using an automated transcription service. The text is indeed extracted in text strings in a subtitle format and a web application is used to link the text with the exact minute of the video or audio recording. These transcriptions, in Spanish and/or Basque, can then be easily used for the judicial review purposes. The training phase of the neural network was made possible with the result of manual annotation of 600 hours of generic recordings, 160 hours of domain content and over 500 documents with domain lexicon (DT4Regions, 2023). With this solution, the manual transcription of recording is no longer necessary, and clerks can now focus their efforts only on correction and completion of transcriptions. 

Expected benefits 

The benefits provided by this solution include: 

• Less workload for clerks in charge to transcribe the judicial trials. The solution enhances productivity in carrying out administrative tasks, making it possible for civil servants to focus only on the revision of transcriptions and save time for other tasks. 

• Less workload for judges who can now consult transcriptions easily and quickly, looking for exact key phrases or names. Therefore, the solution enhances productivity for them as well. 

• Higher interoperability on technical, semantic, and organisational perspective. Automated transcription in a structured format improves data interoperability. This also enhances the legal value of transcriptions, that are now less susceptible to clerical errors.  Lastly, the solution is provided with an application programming interface (API) that can be made available to other organisations. 

Current challenges 

The main challenges related to the development and implementation of this solution were: 

• Poor quality of data needed for the training phase of the neural network. The EJIE needed to sign a contract with an external supplier to ensure quality of the training with additional material. 

• Lack of internal skills. The EJIE needed to involve a private research company to complement the AI-related competencies that were missing in the organisation. 

• Basque Government Informatic Society (EJIE) 

• DT4REGIONS solution description