%%% slides
LDC Software and Annotation Infrastructure
- Systems Group / Systems Service Center formed circa 2008
- Software Group formed 2011 to focus on collection, annotation, delivery
- Hardware concerns shared by both groups, notably:
- Software Group increasing handles VMs in the cloud
- Software Group handles collection equipment, i.e. recording audio and video
Timeline
- Before 2011 siloed desktop apps
- effective but lacked flexibilty
- Since 2011, focus on web apps
- Since 2018, introduction of cloud solutions
- 2025 introduction of Agentic AI
- Future web and desktop apps for secure environments
Annotation Types
- mention, entity, coref: from text extents
- various event and semantic annotations from text, audio, video
- bounding box and sequencing to support multilingual OCR
- audio auditing and waveform based transcription
Customizable Annotation Software
- impossible (ok, impractical)
- Even in house customization only wasn't successful
- As you increase flexibility you increase difficulty
- Technology also changes too rapidly, changing the landscape
- The middle ground of "customizable" really needs developer support
- This is still true with AI agents because they need guidance
To the web and back again
- LDC had web apps in the 90's, the CGI era
- However, serious annotation was desktop until circa 2011
- In a just a few years, all annotation was web based
- The web technology landscape has changed rapidly
- Unfortunately is the change is mostly an increase in complexity
- The move to the web was very much desirable
- However, agentic AI has led me to question if desktop apps should be reconsidered
Links
- This poster describes our transcription tool, but many of the points apply to other tools we've done. asa2021
- This is our PennSound poster. asa2026
- LanguageARC Citizen Science portal. LanguageArc
- Language Game NameThatLanguage
- This shows off some of our recording hardware capabilities. mics
- Our CAMIO paper with pics of bounding boxes camio