Summarizing of parliamentary documents and conversion into podcasts

Transcription and translation

Finland

Use case ID: 044

Author: Parliament of Finland

Date: 8 July 2024 

Objective:

Generate summaries of XML or PDF documents (of between 10 and 1,000 pages) published on the parliamentary website, and create audio files for the podcast. 

Actors:

  • Backend system
  • Application programming interface (API)
  • Parliamentary website
  • Committee members
  • Generative artificial intelligence (AI)-powered assistant
  • Deep learning technology 

Prerequisites:

  • Backend system publishing documents in XML, PDF and CSV formats to the website API
  • API capable of converting documents in predefined formats into HTML and publishing them on the parliamentary website
  • Generative AI-powered assistant and speech learning system with access to the files
  • Generative AI-powered assistant trained on parliamentary material and knowledge and backed up with guidelines, references and citations to source documents, etc.
  • Speech learning tool presets configured 

Scenario:

  1. The backend system publishes documents (e.g. committee reports) to the website API in XML format.
  2. The API converts the XML document into HTML format and publishes it on the parliamentary website.
  3. The API also moves the XML file to the predetermined location.
  4. The generative AI-powered assistant generates a summary from the XML file and saves this in the predetermined location.
  5. The speech learning system transforms the XML file, creating a summary using natural-sounding human speech and saving it as an audio file (MP3, OGG).

Alternate flows:

  • A committee member can choose, from their device, any public parliamentary document published on the parliamentary website for reading or listening.

Expected results:

  • Summaries make it faster and easier to absorb large amounts of material and get an overall picture.
  • Parliamentary materials are more discoverable, accessible and reusable.
  • Listening to the content frees up other senses for other activities (e.g. it is possible to listen to a podcast while walking or driving).

Potential challenges:

  • Gaps in the definitions of AI-tool presets
  • The security of the content, including compliance, privacy and infrastructure
  • The accuracy and truthfulness of the content produced, including potential hallucinations

Data requirements:

  • Public data from parliament

Integrations with other systems:

  • Backend system
  • Integration platform
  • Cloud platform
  • Databases
  • Generative AI-powered assistant 
  • Speech learning system
  • Multi-factor authentication (MFA)

Success metrics:

  • Error rate of no more than 1% in a random sample of summaries
  • Error rate of no more than 2% in a random sample of audio files
  • Web Content Accessibility Guidelines (WCAG) 2.1 Level AA for technical accessibility (score of 80%)
  • Use of parliamentary services measured according to the following three-step scale:

1. Use of the service is technically successful

2. Use is effortless and easy

3. Use is pleasant, interesting and even inspiring

  • Average usability target of level 2

 

The Use cases for AI in parliaments collection is published by the IPU’s Centre for Innovation in Parliament as part of the Parliamentary Data Science Hub’s project to create guidelines for AI governance in parliaments.

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International licence. It may be freely shared and reused with acknowledgement of the author and the IPU. 

A use case describes how a system should work. It is used to plan, develop and measure implementation. A use case is not the same as a case study, which is a descriptive text of an actual project’s implementation. Please note that this use case is provided “as is” and neither the IPU nor the author accepts any responsibility for its use.

For more information about the IPU’s work on artificial intelligence, please visit www.ipu.org/AI or contact [email protected].