The Organization of American States (OAS) is the premier regional forum for political discussion, policy analysis, and decision-making in Western Hemisphere affairs. As such, the OAS Secretariat of the General Assembly, the Meeting of Consultation and the Permanent Council are all mandated to keep minutes of all of their meetings. These meetings are transcribed in order to maximize their value for research and also to reach more diverse audiences.
Previously, this transcription process was manual and time-consuming, requiring personnel to spend up to 20 hours to transcribe a single hour of recorded content. A typical four-hour video, therefore, was taking up to 80 hours for delivery of the completed “Official Document” using the former workflows.
AI-DRIVEN MEDIA ASSET MANAGEMENT TO THE RESCUE
In order to streamline, automate and accelerate the translation process, the OAS has turned to Primestream’s advanced media asset management (MAM) solution, leveraging powerful artificial intelligence (AI) tools and technologies. Based on Primestream’s cloud-based Xchange MAM platform, the new system is expected to deliver significant new efficiencies, cost savings and enhanced service to OAS member countries.
The MAM system transcribes audio content automatically using AI capabilities such as voice recognition, facial recognition and automatic metadata tagging. Next, the system outputs the transcription in multiple languages for rapid delivery to delegates from the 35 OAS member countries. The tagged transcriptions are uploaded to Xchange’s centralized, cloud-based storage to make the media content searchable and give OAS staffers fast and easy access to all media assets. For instance, they are able to search on representative speakers whose content has been tagged using facial/voice recognition and metadata.
A CLOSER LOOK AT THE OAS CONTENT WORKFLOW
At the OAS, media users and metadata organizers are tasked with creating metadata descriptions and uploading the completed recordings of the events to the Primestream Xchange cloud storage environment. There, the content is associated with the descriptive metadata relating to the event, such as date, name, location and agenda, and stored in a “meetings” placeholder. Next, the MAM system automatically creates a proxy and sends all uploaded clips to the Azure Video Indexer platform for AI analysis of the content.
At this stage, the facial recognition module within the Xchange Elastic Data Viewer comes into play. By creating an index of faces appearing in each video, the system is able to catalog and segment the content by participant. With indexed markers pointing to specific ambassadors and other officials who take part in multinational meetings, the Elastic Data Viewer makes it easier for staffers to access content based on meeting participants.
The uploaded clips are then transcribed in the four official languages used in OAS meetings—English, French, Spanish and Portuguese. When the AI engine completes its analysis of the uploaded content, an email is automatically generated to notify the other user groups that there is a clip ready to be reviewed.
OAS editors then review each clip using the Xchange Elastic Data Viewer, which offers a timeline for displaying the data from the AI engine analysis. The editor reviews the analyzed content to confirm the correct meeting has been uploaded and to add any corrections or additions to the facial recognition done by the AI analysis. Changes and additions to faces help the AI engine “learn” to identify the same faces in the future.
Once the editor’s job is complete, a transcriptor reviews and corrects the transcript from the audio of the analyzed content. Finally, an approver gives final signoff that the corrected transcripts and other analyzed metadata are accurate. It’s the approver’s job to create the “Official Document” in Word format and then upload and attach it to the associated meeting video.
With the Official Document uploaded, end users are now able to search on the content by name, date and location and access the “Official Document” of the meeting with associated video and audio assets.
The OAS’ automated, AI-driven transcription process based on Primestream Xchange is a major improvement over the previous manual process, which could take up to 30 days to accomplish the same tasks. Now, instead of the 80-hour turnaround to produce an Official Document for a four-hour video, the multilanguage transcription automation of Xchange is able to process the same video in eight hours—a 90% time savings.
By accelerating cataloging and transcription delivery, the OAS is now able to serve its members in a way that would not be possible without the adoption of AI and dynamic media asset management capabilities. With the Xchange platform deployed in a cloud-based architecture, the OAS will be able to extend access easily to its video content and related documentation to its offices across the Americas.
This article was written by Claudio Lisman, president and CEO of Primestream