Design, Development, and Implementation of AI tools for transcribing Court proceedings and arguments: Supreme Court

On May 23, 2023, the Supreme Court of India invited bids to deploy Artificial Intelligence tools for the transcription of oral arguments and court proceedings. Earlier in a Court proceeding where the Constitution bench headed by Chief Justice of India DY Chandrachud was hearing a matter, AI was used to transcribe the court proceedings and arguments in real-time. The transcriptions were shared with the concerned lawyers, then uploaded on the official website of the Supreme Court. This helps in increasing accountability within the top Court. In this context, the Supreme Court invited bids in three-cover systems including EMD, Financial Bid, and Technical Bid, for the Design, Development, and Implementation of AI tools for transcribing proceedings and arguments. A bid document illustrated that “The AI tool for transcription of arguments and court proceedings shall initially be deployed for the Constitution Bench matters and with the passage of time extendable to all matters, subject to the approval and directions of the concerned Court.”

Moreover, a bid document highlights the working process of transcription, “The transcription of arguments generated through AI, Machine Learning, and Deep Learning solution and tools must be proof-read by qualified transcribers as the first line and thereafter by reviewers, who must proofread the same and remove any typographical errors, mistakes before forwarding the final version transcription to the Registry. The Supreme Court intends to publish the AI-generated transcription on its official website and to share the same with the concerned Advocates through electronic mode.” The transcriptions should be available in easily accessible file formats on the computer’s operating system and other electronic devices. A bid document added, “The transcription generated must be in searchable format, PDF/A (for archiving and long-term preservation), HTML, Word Process, etc. The transcription generated through AI tools must bear time stamp and be capable of integration with video recordings of court proceedings. The AI tool to be deployed must have all the latest features, particularly bookmarking, labels, and makers for quick access and convenience. The AI tool must be capable of generating metadata and have a mechanism to share the transcription either through an Application Programming Interface (API) or any other latest mechanism as per the industry standards.”

Also, the AI tools should be capable of generating summarized transcription, be it speaker centric and/or party specific, etc. for further classification. A bid document further stated that “The AI solution and tools will be hosted on the server(s) of the Supreme Court of India, which can be on-premises and/or cloud, and if the solution is server-less, then the same shall remain within the control of the Supreme Court of India. AI tools shall not use any external APIs, tools, or packages developed by third parties, which are not free and open source, to avoid any dependency on its functioning.”