Enable transcripts for video content

Owned by Amol Ghatol
Nov 10, 2021
3 min read

Introduction

This wiki explains the design and implementation of the enable transcriptions for video content.

https://project-sunbird.atlassian.net/browse/SB-26446

Key Design Problems:

  1. How to store transcripts files in the system.

  2. Support configuration for transcript file types.

  3. Relation between transcripts files & content.

Solution:

  • We want to create transcript files as assets in our system. For that, we want to create a new asset primaryCategory with mime-types which are supported to create transcript files.

primary category: Video Transcript

tragetObjectType: Asset

mimeType Supported: application/x-subrip

Currently, we are supporting only a single mimeType. in future, if we want to support multiple mimeType just need to update the primary category definition.

{ "objectCategoryDefinition": { "categoryId": "obj-cat:video-transcript", "targetObjectType": "Asset", "objectMetadata": { "config": {}, "schema": { "properties": { "mimeType": { "type": "string", "enum": [ "application/x-subrip" ], "default": "application/x-subrip" } } } } } }

  • We are introducing new metadata property transcripts in the content object. this property will have a list of transcripts files uploaded by users for specific content.

Schema changes:

{ "transcripts": { "type": "array", "items": { "type": "object" } } }

Sample: List of transcripts files uploaded by users for specific content

{ "result": { "content": { "identifier": "do_12345678", "name": "Video Content", "author": "N118",, "transcripts": [ { "language": "English", "languageCode": "en", "identifier": "do_1234", "artifactUrl": "https://dockstorage.blob.core.windows.net/sunbird-content-dock/content/do_11340589048704204811789/artifact/do_11340589048704204811789_1636461247932_sample.srt" }, { "language": "Hindi", "languageCode": "hi", "identifier": "do_5678", "artifactUrl": "https://dockstorage.blob.core.windows.net/sunbird-content-dock/content/do_11340589048704204811789/artifact/do_11340589048704204811789_1636461247932_sample.srt" }, ] } } }

  • We are not maintaining any relations between content & transcripts(assets). we are adding a list of transcripts as metadata to the content.

  • Transcripts property we are indexing into elastic search as an object. So it will be easy to filter contents based on transcripts language & identifier