Releases: AI4Bharat/Shoonya
Shoonya v3.0
This release adds on following new features on top of Shoonya v2.4 :
- Flower Configuration for asynchronous celery tasks logging.
- Deletion confirmation endpoint.
- Added new notification types.
- Access level changes for manager role.
- Minor changes to analytics.
- Added new project types - OCRSegmentationCategorization & editing
- Dataset and Task data privacy changes to hide public URLs.
- New endpoint to add a proxy Audio URL service.
- Line Charts based performance analytics.
- Integration changes to AcousticNormalizedTranscriptionEditing project type.
- Exception handling and code refactoring for backend codebase.
- Download endpoints optimisations.
- Endpoint support to store transliteration logs to blob storage.
- User active/inactive integration.
- Decentralized User Invitation to workspace managers.
- Changes to email templates.
Shoonya-master.zip
Shoonya v2.4
Introducing Shoonya v2.4 with following new features:
- Workspace and Organisation level Payment Reports.
- Support for OCR, ASR data types predictions population.
- Consider Batch sampling and automatic annotations creation support to pull new data items.
- Reports scheduling feature.
- Backend support for Chitralekha UI for any Audio Project Types.
- New project types Acoustic Normalized Transcription and Acoustic Normalized Transcription Editing utilising CL UI.
- Transliteration Logging support using Blob storage.
- Support e-mail-based async calculation to all reports.
- Code Refactoring and reformatting.
- User profile picture upload and change feature using blob storage.
- Elastic search and Kibana support for logging setup.
- Download all projects within a workspace asynchronously using blob storage.
- Some bug fixes for annotation filter, reports mail along with others.
Shoonya v2.3
Introducing Shoonya v2.3 with powerful features :
- Conversation Verification Project Type.
- OCR Project Types modifications.
- Option to change the stage of a project to Supercheck Stage.
- All endpoints for Supercheck workflow.
- Bug fixes for Assigning and Unassigning Tasks.
- New field domain in TranslationPair datatype.
- Superchecker notes.
- Reports Bug fixes.
- Frozen Users for Workspace.
- Re-invite users.
- Modification for Login and Change Password.
- Automatic Annotation Creation for external data.
- Intra-dataset Automation to populate draft_data_json.
- More quality parameters in reports (WER, segment length etc).
- Workspace-level analytics.
Shoonya v2.1
Shoonya v2.1 has some considerable changes, in addition to the introduction of few new roles :
- New user roles for Reviewer, Super-checker, and Admin.
- Updated endpoints relevant to Reviewer and Admin roles.
- Integration with new Indic-Trans-v2 deployed on Dhruva.
- New field in the project model to support project stage in place of deprecated field review enabled.
- New annotation and task statuses to support super-checker flow in upcoming versions.
- annotation_type field in all annotations to signify which role it belongs to.
- New field named revision_loop_count in the task model to support bookkeeping.
- New field named super_checker_user in the task model.
- Changes to migrate user roles based on the appropriate work they are assigned to.
Shoonya v2.0
Shoonya v2.0 has some major changes, in addition to the introduction of new project types support.
- Design changes to have a status associated with each annotation
- New project Types support for:
- Domain Classification along with Sentence verification.
- Audio Segmentation
- Audio Transcription Editing with support for populating predictions.
- Glossary support on annotation page of Translation Projects.
- Support for reviewers to accept a task with major or minor changes.
- Draft and skip option for reviewers.
- Tags support to allow noise tagging in all Audio project types.
- Improved UI to increase Audio Transcription productivity.
- Word count/ Audio duration based public analytics.
- Optimized project listing based on recently worked project.
- Export fix in conversation translation editing project type.
- Support search and filter based task flow for Start Labeling Now button
- New task status to reflect whether a task is exported.
- All tasks tab for managerial view of a project
- Frontend bug fixes for Automate Datasets page.
- Filters for Projects listing and Datasets listing pages.
- Support to download all annotations from all tasks of a Translation Project.
- Endpoint to allow managers to deallocate tasks for any user in the project.
Shoonya v1.3
Shoonya v1.3 focuses on new features related to reports, along with a new Project type support Audio Transcription Editing
Features in this release are listed below:
- New project type for Single speaker transcription editing.
- Public API endpoint for language based Organization Analytics.
- Complete support for all levels of review reports.
- Annotation Quality Reports.
- Backend support to categorize accepted with major/minor changes.
- Search support for Dataset Items table.
- Bulk delete endpoints for tasks and data items.
- Support for Managers, Org Owners to be able to annotate tasks.
Shoonya v1.2 Release
Shoonya v1.2 focuses on new features along with Project type supporting Conversation Translation Editing.
Features in this release are listed below:
- Improved User reports for annotations and review.
- Patch to update conversation data type to support Machine Translations.
- Celery-based implementation for Automated MT function for Conversation Data Type.
- Removal of task-lock deprecated functionality.
- Workspace-level User analytics for review workflow.
- Endpoint to support bulk deletion of data items and all linked tasks.
- Organization-level User analytics for review workflow.
- Review workflow-based reports for the Analytics tab (publicly accessible endpoint).
- Review reports for User-level progress.
- Integration with Azure Translate.
- TSV support for Projects download.
- Endpoint to support Normalized character-level edit distance between sentences.
- Support to filter by task status while downloading projects.
- New Boolean field in the Users model to support user input for receiving daily mails.
- Endpoint and celery-beat setup for sending daily progress emails to Users
- Support for annotation, and review reports for daily progress mails.
Shoonya v1.1
This release focuses on a few bug fixes, and covers some user requests, along with dataset automation functions mentioned in detail below:
- Rename Task status
rejected
toto_be_revised
. - Fix a bug in the review feature.
- Add reviewer reports for project analytics.
- Refactor
User
field toAnnotator
in Project model. - Functionality to remove users from workspace.
- Code refactoring to take
user_id
instead ofusername
oremail
in various endpoints. - Support automated transformation of Sentence Text Datasets to Translation Pairs Dataset involving functions for Generating Machine Translations using IndicTrans and Google Translate models.
Shoonya v1.0
Shoonya v1.0 supports the following features :
- Supports all the 22 official Indian languages
- Currently support Sentence Verification tasks, Context Translation Verification project types
- Provides AI support with translation
- Cleaner hierarchy of Organization, Workspace, Projects.
- Reports at various levels (Org, workspace, project, user) and multiple dimensions
- Allow creation of task chains and custom inputs as required by Language Experts
- Enables language coordinators to enable effective collaboration (Shareable Notes, Drafts)
- RTL and Transliteration based support