Joseph Lee

@[email protected]
61 Followers
4 Following
179 Posts
A researcher+grad student during the week, a tech observer+accessibility advocate in the weekends. A blind 1.5-gen Korean American graduate student working at the intersection of organizational communication, technology, and disability studies, I constantly look for ways to learn about how people and organizations define, discuss, and practice accessibility.
Certified NVDA Expert, member of Cal State LA, CU Boulder, NVDA screen reader, and Windows Insider communities
PronounsHe/his/him
KeywordsOrganizational communication, communicating technology design, disability studies and accessibility
Languages (human and computing)English, Korean, Python, C++
Show threadJoseph Lee5d ago
Further, the pull request I linked to at the start of this posts chain had a major issue: the command supposedly assigned to recognie text content using an on-device model is actually unavailable; NVDA+Shift+R is NOT a global command but is defined in Excel app module to mark row/column headers for spreadsheets. Therefore, the AI must parse the entire NVDA source code next time (this takes time and resources). 4/
Show threadJoseph Lee5d ago
Pull request templates exist for a number of reasons, including communicating expectations by the organization/project and to help pull request writers (humans or machines) organize thoughts and information in a structured way. In organizatoinal communication space, this means when an AI generates a PR and the template content, it is communicating the knowledge gathered by an organization that can conflict with expectations from another organization. 3/
Show threadJoseph Lee5d ago
In recent days, several NVDA screen reader pull requests were closed because NV Access realized that these PR's were written by AI without filling out the NVDA PR template:
https://github.com/nvaccess/nvda/pull/19774
https://github.com/nvaccess/nvda/pull/19775
https://github.com/nvaccess/nvda/pull/19776
2/
fix: guard oldTracker.pluralTimeout before comparison in makeMergedTrackerIfPossible by seang1121 · Pull Request #19774 · nvaccess/nvda

Fixes #19620 Summary In makeMergedTrackerIfPossible, the pluralization branch checked newTracker.pluralTimeout is not None before using it, but did not apply the same guard to oldTracker.pluralTime...

GitHub
Joseph Lee5d ago
To #NVDASR community: please, PLEASE don't just use an AI system to submit pull requests:
https://github.com/nvaccess/nvda/pull/19787
1/
Joseph LeeFeb 21
Christopher Duffley - KC1WXPFeb 21
@joslee and I are delighted to announce the release of StationPlaylist add-on 26.03!
Changes:
* Initial support for StationPlaylist Recorder (not to be confused with VT Recorder).
* NVDA will announce labels for SPL Recorder configurationnnnnnn screen controls.
* Most if not all multicolumn lists now have column explorer and column navigation support. Mainly tested in local Studio's Listener Stats window and the same list in Options > Listener Stats.
Download from NVDA's Add-On Store and also at https://github.com/ChrisDuffley/stationPlaylist/releases/download/26.03/stationPlaylist-26.03.nvda-addon
Enjoy!
Joseph LeeFeb 7
Guide: porting NVDA add-ons to 64-bit NVDA (2026.1): https://nvda-addons.groups.io/g/nvda-addons/topic/guide_how_to_port_add_ons_to/117685297
TL:DR: check the add-on code, look for 64-bit dependencies if needed, test the add-ons, then publish it (hopefully to the add-on store).
Reason: add-ons are crucial part of the NVDA experience. Porting add-ons to 64-bit NVDA keeps them and NVDA relevant. The greatest thing that can happen to the community in 2026 is discovering the joys of using add-ons once again in the midst of celebrating NVDA's 20th anniversary.
Show threadJoseph LeeJan 31
FAQ 4: is BNE using proprietary braille code for input and output?
Answer: NVDA uses liblouis (open-source) braille translation module. This means the available braille tables will depend on which liblouis version is used at a given moment (remember, NVDA receives updates several times a year). 7/
Show threadJoseph LeeJan 31
FAQ 3: can you turn off speech on the BNE?
Yes in two ways: one temporary, one permanent.
* Temporary: NVDA let you cycle between four speech output modes: talk, beep, on-demand (similar to speech on request mode), and off. It is also possible to exclude certain speech output modes.
* Permanent: NVDA does have a "no speech" speech synthesizer option. 6/
Show threadJoseph LeeJan 31
FAQ2 cont: GitHub issues:
* Braille is not reporting Excel cell column headers: https://github.com/nvaccess/nvda/issues/6901
* Reporting column headers before content in speech: https://github.com/nvaccess/nvda/issues/4629
The first issue is a bit geeky as it discusses some technical background. As this is an implementation discussion for NVDA, HumanWare may not be the right entity to resolve this (this is a cross-device problem, not unique to BNE). 5/
Show threadJoseph LeeJan 31
FAQ 2: in Excel, can you change the column header/content annoucement? Curretly NVDA reorts cell content, cell coordinates (if told to do so), and then column header (column heades reporting must be configured).
Answer: not without additional work. Currently NVDA is not brailling Excel cell column headers, so this must be resolved (issue is open on GitHub). Once this is done, a setting may need to be added to swap column header/content announcement. 4/