Attachment 1 - Required Capabilities_ESR STT RFI.pdf

Electronic Sound Recording with Speech-to-Text Solution

REQUEST FOR INFORMATION

THE ADMINISTRATIVE OFFICE OF THE UNITED STATES COURTS IS

SEEKING SOURCES FOR:

A vendor to provide an electronic sound recording product with a speech-to-text technological solution.

In September 1999, the Judicial Conference of the United States approved electronic sound recording (ESR) technology as a method of taking the official record in federal courts. The U.S. Courts are currently using ESR from For The Record, Liberty, and CourtSmart vendors as their recording technology within the courtrooms.

INTRODUCTION:

The Administrative Office (AO) has been made aware by numerous district courts that there is increasing difficulty recruiting qualified court reporters across the federal judiciary. The COVID-19 pandemic has exacerbated these recruitment issues. To contend against this growing trend, many district courts have increased electronic sound recording (ESR) equipment use.

The Court Services Office (CSO) in the Department of Program Services (DPS), in partnership with the Technology Solutions Office (TSO) in the Department of Technology Services (DTS) of the AO, is seeking a solution that:

• Utilizes the existing court-owned audiovisual (AV) components.

• Records all audio sources in a multi-channel audio file.

• Provides court with note-taking option.

• Converts live court proceeding audio sources (English only) into text-based output

(i.e., text/caption file format only - not as an official transcript).

REQUIRED CAPABILITIES:

Functional

The AO is looking for a solution to digitally record audio and provide speech-to-text transcription during court proceedings (such as hearings and trials) directly from a courtroom AV system. The solution would be for all court types: Appellate, District, and

Bankruptcy. There are approximately 2,200 courtrooms that this product may be utilized in.

The solution shall record multiple audio channels and playback the court proceedings.

The solution shall have a method of labeling audio channels (e.g., judge name(s), prosecution, defense, witness, defendant, plaintiff, lectern, juror box, debtor, etc.). Each channel’s audio shall be separately manageable, able to mute/unmute individual channels in any combination, and able to adjust each channel's volume (increase/decrease). Notes and speech-to-text conversion shall have time stamps when participants speak. The audio shall be recorded and saved to a local internal or external hard drive while also archiving the audio to a network location. Separate from the standard saving and archiving processes, an automated National Archives and Records Administration (NARA) compliant format shall be archived in a separate NARA folder for NARA retention schedules. A single audio recording shall continue as long as the solution operator has the recording process activated.

The solution shall provide staff with a built-in suite of note-taking and editing tools, including keyword searching of text to locate words or phrases. Notes created shall attach a time stamp with a link to the audio at the specified time.

The solution shall perform speech-to-text conversion using a common English language dictionary with at least a 90% accuracy rate. The solution operator shall see the text on-screen as it is converted. The judge shall see the speech-to-text in a separate client on the judge’s computer or tablet device. The solution shall provide for scrolling and searching text within the speech-to-text viewing client, allowing the judge and court staff to take private notes. As the spoken word is converted to text, the solution shall automatically insert tags with a link every time the speaker changes to allow playback of the audio file from that respective time-stamped location. The solution shall provide the ability to read or playback any portion of the recording at any time during a proceeding.

Technical

• General o Shall function on a Windows-based Operating System, version 10 or newer, on a PC-based computer or laptop.

The additional ability to operate within a virtual desktop infrastructure (VDI) environment is preferable.

o The solution shall have a user-friendly interface that provides easy access to the intended separate functions for recording, playback, and note-taking for the solution operator.

o A controlled process for software updates shall be included in the solution.

o The solution should allow for security and functionality updates on a scheduled basis.

o The solution shall have user access controls to grant or deny access to individual modules within the solution. The user may be authenticated via Microsoft Active Directory.

o The solution shall have user access controls to remote clients or devices connecting to receive a near-real time stream of the proceeding. Users for this feature may not be in the Active Directory, and guest accounts established by the court locally in the solution.

o The solution shall have reporting capabilities to report, at a minimum:

Time in court by a judge, Time in a courtroom by any judge, Time in court by the solution’s operator (user logged into Windows).

• National Archives and Records Administration (NARA) Compliance o The solution shall save the audio in a preferred NARA digital format.

o The solution shall conform to the audio file naming guidance provided in the NARA Bulletin 2015-04 Appendix B.

o The solution shall conform to any NARA requirements concerning metadata attached to the audio file. See NARA Bulletin 2015-04.

• ESR Recording & Note Taking Client o The solution shall store the audio as a digital file.

o Separate from the manufacturer’s proprietary digital file format, the solution shall include an automated process that:

Delineates the beginning and end of all recordings.

Includes a quick-key or functional button to delineate the beginning and end of a case for continuous recording of cases.

Associate the recordings by date and unique case number.

Separate of the manufacturer’s proprietary audio file / audio wrapper file format, saves the proceedings audio to the required NARA-compliant format.

Separate of the manufacturer’s proprietary note file format, saves the proceeding’s non-private notes to the required NARA-compliant format.

Bundles the NARA-compliant audio and note files for record retention requirements.

o Shall digitally process all courtroom audio (e.g., spoken word, video conferencing, telephone conferencing audio, and/or evidence audio) and may include mechanisms to boost or clarify the audio signal or to save the recording in a compressed format.

o Shall provide confidence monitoring:

https://www.archives.gov/records-mgmt/policy/transfer-guidance-tables.html#digitalaudio https://www.archives.gov/records-mgmt/bulletins/2015/2015-04-appendix-b.html https://www.archives.gov/records-mgmt/bulletins/2015/2015-04.html

To ensure the courtroom audio has been recorded accurately and stored to media.

Shall include safeguards to prevent accidental erasure or over-recording of the court proceedings and provide automatic error detection to ensure continuous, uninterrupted recording for clear playback and transcription.

Shall include an output for off-media monitoring.

o Shall provide a multi-channel recording solution.

The recording solution must include a minimum of eight channels of audio.

Shall allow for the modification of audio channel labels (e.g., judge name(s), prosecution, defense, witness, defendant, plaintiff, lectern, juror box, debtor, etc.).

o Shall provide a note-taking option:

The solution shall create a time stamp synced to the audio recording when the note is created.

Notes shall be searchable.

A hotkey function to allow pre-created phrases to be automatically inserted into the notes at a user-defined location. This is user-customizable depending on who is logged in.

o Shall have a Sealed status for recordings and associated public notes that:

Shall require a password to be entered.

Shall have an administrative solution to recover the password.

Shall allow setting the sealed status before recording.

Shall allow toggling off/on the sealed status during a recording, explicitly allowing a portion of the recording to be sealed.

Shall allow setting the sealed status after a recording.

Shall allow selecting a portion of a recording as sealed after a recording.

o Shall have a Highly Sensitive Document status for recordings that:

Shall record to a thumb drive separate from the standard file location.

Shall require a password to be entered.

Shall have an administrative solution to recover the password.

Shall allow setting the sealed status before recording.

Shall allow toggling off/on the sealed status during a recording, explicitly allowing a portion of the recording to be sealed.

Shall allow setting the sealed status after a recording.

Shall allow selecting a portion of a recording as sealed after a recording.

• ESR Playback Client o The solution shall provide the capability to playback the current or previous proceedings’ audio through the courtroom AV system.

o During the playback of the current or previous proceedings’ audio, the solution shall continue to record the current proceeding’s audio. This should be considered as if it were evidence being played from an audio source within the courtroom or from a remote source such as a telephone or video conference.

o The solution shall provide channel isolation with the ability to:

Adjust (increase/decrease) the audio channel’s volume level.

Mute the individual channel or all channels’ audio.

Adjust (increase/decrease) the playback speed.

• Speech-to-Text with Private Notes o As applicable, the vendor shall identify all data centers where the data at rest or data backup will reside. All data centers shall be guaranteed to reside within the United States. The solution shall perform near-realtime conversion of the spoken word utilizing a common English language dictionary with at least a 90% accuracy rate.

o The solution shall perform near-realtime conversion of the spoken word utilizing a common English language dictionary with at least a 90% accuracy rate.

o The solution shall learn speech patterns and dialects and improve speech-to-text accuracy.

An option allowing for additional court-specific dictionaries to be added as necessary by the courts would be beneficial in the creation of industry-specific language or jargon.

o The solution’s speech-to-text output shall be saved with the digital audio and notes files.

o The solution shall provide for scrolling and searching of speech-to-text output text.

o The speech-to-text conversion client of the solution should function on a laptop or desktop computer with a Windows-based Operating System, version 10 or newer, on a PC-based system, separately from the ESR solution. A website or local host web server with user access control is permissible to allow access from tablet devices.

o The solution shall provide for private notes within the speech-to-text viewing client.

o Private notes shall be saved separately from the audio recordings and ESR recording solution notes location.

o Private notes shall require a password to be entered.

o Private notes shall have an administrative solution to recover the password.

o Private notes shall not be included in the NARA-compliant files.

o The solution shall identify the party speaking in the speech-to-text output.

• Communications o The solution shall provide for Real-Time Messaging Protocol (RTMP) streaming.

The solution shall default to audio streaming to be off.

The solution shall allow the court to turn streaming on or off at any time during a proceeding.

The solution shall provide audio RTMP streaming.

o The solution shall provide interpretation channels (E.g., foreign language, sign language).

The solution shall allow the interpreter to do simultaneous interpretation without interruption to the court proceedings.

Interpreters must be able to be assigned to interpretation channels and the main ‘hearing room’ so that they may hear all court and evidence audio, with the ability to speak to both the court and the parties needing interpretation separately.

Interpreters must be able to switch audio from the interpretation channels to the main hearing room without court intervention.

The solution shall provide for multiple interpreters should there be a need for:

• Team interpretation.

• Multiple simultaneous and separate languages and/or sign language.

The ability to record the interpretation channel should be a controllable option the court can turn on and off (default is off) during a hearing.

SECURITY:

Data Protection

Regardless of solution deployment type, data requiring protections related to encryption, both in transit and at rest and as stipulated in the Technical section of this document, need to address the implementation of cryptographic protections. Technical solutions sought by the Judiciary prioritize conformance with Federal Information Processing Standard (FIPS) Publication 140-2, Security Requirements for Cryptographic Modules.

Web Application Protections https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.140-2.pdf https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.140-2.pdf

Regardless of the deployment type, proposed solutions utilizing web applications must address the utilization of the Open Worldwide Application Security Project (OWASP) best practices. Technical solutions sought by the Judiciary prioritize conformance to at least an OWASP Application Security Verification Standard (ASVS) Level 2.

Compliance

Regardless of solution deployment type, proposed solutions need to address documentation of security mechanism implementations consistent with technical security controls in the Judiciary Information Security Framework (JISF), and must include the ability for individual Judiciary organizational entities to configure the solution’s security mechanisms sufficient to satisfy unique JISF organizational parameters.

Libraries, Frameworks, and Products

Regardless of solution deployment type, the vendor shall disclose in writing to the Judiciary all third-party software used in the solution, including all libraries, frameworks, components, and other products, whether commercial, free, open-source, or closed-source. The Judiciary seeks a vendor to make reasonable efforts to ensure that third-party software meets all the security requirements and is as secure as custom-developed code.

Artificial Intelligence

Regardless of deployment type, and as applicable to technologies using artificial intelligence, the vendor shall disclose in writing to the Judiciary, and in alignment with guidance related to NIST AI 100-1, Artificial Intelligence Risk Management Framework (AI RMF 1.0), any considerations or assumptions related to applicable developer activities included in the Planning and Design; Collect and Process Data, Build and Use Model; and Verify and Validate phases of the AI RMF Lifecycle.

ADDITIONAL REQUIREMENTS:

The currently installed electronic sound recording systems used in federal courtrooms provide continuous and uninterrupted recording for clear playback and transcription. The solution shall be efficient for all users by increasing accuracy and reducing labor costs.

All audio formats shall be non-proprietary, contain open-source coding and technology, and comply with NARA's digital audio and digital video file format guidance.

Additionally, the solution must allow for the audio files to be accessible and secured confidentially on the federal judiciary’s network infrastructure, either locally hosted or cloud-based.

https://owasp.org/www-project-application-security-verification-standard/ https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf