MiniConf: A Virtual Conference in a Box(mini-conf.github.io)
mini-conf.github.io
MiniConf: A Virtual Conference in a Box
https://mini-conf.github.io/index.html
24 comments
This is both usable and has a unique "vibe", in a time when many virtual events have struggled to capture the character and human essence of their F2F events. Congratulations and thank you for releasing the code!
A few production questions.
1. Sync/playback is relatively fast. Are you using a streaming server or web-hosted video files?
2. On pages with video & slides, how are the two frames synchronized, e.g. are those two synchronized video streams with known timecodes for slide changes, or is the slide frame showing static images that are synced with video timecodes?
3. Did you chose Python because of existing code that you wanted to reuse?
4. Which existing software was closest to meeting your needs, and what were the shortcomings that motivated miniconf?
A few production questions.
1. Sync/playback is relatively fast. Are you using a streaming server or web-hosted video files?
2. On pages with video & slides, how are the two frames synchronized, e.g. are those two synchronized video streams with known timecodes for slide changes, or is the slide frame showing static images that are synced with video timecodes?
3. Did you chose Python because of existing code that you wanted to reuse?
4. Which existing software was closest to meeting your needs, and what were the shortcomings that motivated miniconf?
Thanks! If you are interested more about the event here was a podcast https://www.thetalkingmachines.com/episodes/iclr-accessible-... and blog post https://medium.com/@iclr_conf/gone-virtual-lessons-from-iclr...
Answers:
1 + 2. We used https://slideslive.com/ an external conference provider. They do a lot of optimizing behind the scenes, and have that cool sync.
3. Python is the de facto standard language in ML. We use it for everything. We thought about js/react or something, but we didn't think it would be maintainable.
4. A lot of the existing software assumed there would be a physical event or even worse tried to pretend there was a virtual physical event. We wanted something that was async-first, desktop-oriented, and very simple to browse.
Answers:
1 + 2. We used https://slideslive.com/ an external conference provider. They do a lot of optimizing behind the scenes, and have that cool sync.
3. Python is the de facto standard language in ML. We use it for everything. We thought about js/react or something, but we didn't think it would be maintainable.
4. A lot of the existing software assumed there would be a physical event or even worse tried to pretend there was a virtual physical event. We wanted something that was async-first, desktop-oriented, and very simple to browse.
So MiniConf is effectively a framework for maintaining a conference/event website. It takes care of presenting the program, schedule and ties together the presentations content (video, slides) which are hosted and served by third-party providers.
The result is a static website where the participants meet at time, but all interactions are handled by the third-party providers.
The main features are ease of configuration (CSV files) and templatability (with Flask).
Is my understanding correct?
I was wondering if it offered similar features to EasyChair, like participant registration, review process, notifications etc.
The result is a static website where the participants meet at time, but all interactions are handled by the third-party providers.
The main features are ease of configuration (CSV files) and templatability (with Flask).
Is my understanding correct?
I was wondering if it offered similar features to EasyChair, like participant registration, review process, notifications etc.
Yup that's right. Although I would add front-end search and visualization (with d3) to the features.
As you mentioned there are a lot of good solutions for review, notification, and reg. Common choices are OpenReview/CMT/HotCRP/EasyChair/softconf... They're all pretty mature and complex.
The thing that people suddenly need is ways to easily customize their own pages for hosting virtual conferences and workshops. That's the problem miniconf targets.
As you mentioned there are a lot of good solutions for review, notification, and reg. Common choices are OpenReview/CMT/HotCRP/EasyChair/softconf... They're all pretty mature and complex.
The thing that people suddenly need is ways to easily customize their own pages for hosting virtual conferences and workshops. That's the problem miniconf targets.
For people looking for a Free Software solution for running a (traditional) conference, there is also PreTalx https://pretalx.com/p/about/ and https://github.com/pretalx/pretalx
But your simple solution for virtual events is very much welcome and appreciated! Many thanks!
But your simple solution for virtual events is very much welcome and appreciated! Many thanks!
Thank you for cleaning up the improvisational work you did for your team’s own needs and making this available to everyone.
The mini-conf tour is a good overview for generalized application. Your ICLR pages are great examples of mini-conf in action, e.g.
https://iclr.cc/virtual_2020/workshops_8.html https://iclr.cc/virtual_2020/papers.html?filter=keywords
The mini-conf tour is a good overview for generalized application. Your ICLR pages are great examples of mini-conf in action, e.g.
https://iclr.cc/virtual_2020/workshops_8.html https://iclr.cc/virtual_2020/papers.html?filter=keywords
Yeah :) we hacked in a lot of things last minute for ICLR. MiniConf is a clean 80% functionality / 20% code version of that codebase.
I love this. Conferences and the publishing business are extremely scammy IMO. There's no reason you need to pay anyone $500 to host a pdf.
What researchers want is discoverability and credit. Some researchers are already famous and own their own distribution via arxiv or Twitter but for new researchers publishing a paper at a prestigious conference is a way to bootstrap a reputation.
If you really wanna get rid of large publishers what you really need to solve is discoverability and you can solve it with a combination of virtual mini conferences and callouts on Twitter and Hacker News.
I've gotten some of the best constructive feedback on my blog posts from Hacker News directly, I don't see how that's different from peer review. I don't find it very likely that someone would even be willing to give me feedback unless they themselves are already subject matter experts.
But really the impact of a piece of computer science research is directly correlated to how many people directly use the OSS inspired by the research. People in industry know this, what's left is is for tenure committees to also take software impact into account.
What researchers want is discoverability and credit. Some researchers are already famous and own their own distribution via arxiv or Twitter but for new researchers publishing a paper at a prestigious conference is a way to bootstrap a reputation.
If you really wanna get rid of large publishers what you really need to solve is discoverability and you can solve it with a combination of virtual mini conferences and callouts on Twitter and Hacker News.
I've gotten some of the best constructive feedback on my blog posts from Hacker News directly, I don't see how that's different from peer review. I don't find it very likely that someone would even be willing to give me feedback unless they themselves are already subject matter experts.
But really the impact of a piece of computer science research is directly correlated to how many people directly use the OSS inspired by the research. People in industry know this, what's left is is for tenure committees to also take software impact into account.
> I don't see how that's different from peer review
A lot of people think they're subject-matter experts and may comment on your article, but may be mistaken.
In all the peer review processes I've been involved in reviewers review each others reviews before they go out. Other reviewers will challenge a bad or misleading review. If there aren't people who know enough about the subject they'll go out and find someone who is.
You could literally do this process in a web interface that looked like Hacker News, but they already pretty much do.
I think if you built your dream review system... I think you may find it looks pretty much like what we have already.
> callouts on Twitter and Hacker News
Sounds like a popularity contest.
A lot of people think they're subject-matter experts and may comment on your article, but may be mistaken.
In all the peer review processes I've been involved in reviewers review each others reviews before they go out. Other reviewers will challenge a bad or misleading review. If there aren't people who know enough about the subject they'll go out and find someone who is.
You could literally do this process in a web interface that looked like Hacker News, but they already pretty much do.
I think if you built your dream review system... I think you may find it looks pretty much like what we have already.
> callouts on Twitter and Hacker News
Sounds like a popularity contest.
We’re building a platform for crowdsourcing the research process, from ideation to peer review at https://asone.ai.
It’s kind of like arXiv with voting and comments, but also supports collaboration for research in progress.
> You could literally do this process in a web interface that looked like Hacker News, but they already pretty much do.
I would love to hear your thoughts about how the current systems might be improved and if we’re on the right track.
It’s kind of like arXiv with voting and comments, but also supports collaboration for research in progress.
> You could literally do this process in a web interface that looked like Hacker News, but they already pretty much do.
I would love to hear your thoughts about how the current systems might be improved and if we’re on the right track.
> People in industry know this, what's left is is for tenure committees to also take software impact into account.
Software should be citable. You can make it so by publishing a paper describing the software. Usually the authors then request anybody who uses the software in their research to cite that paper.
Software should be citable. You can make it so by publishing a paper describing the software. Usually the authors then request anybody who uses the software in their research to cite that paper.
You can publish the code itself somewhere like zenodo or figshare (disclaimer, I work for Digital Science) and have it citable like that. No need to publish a paper on it. This is how I've published code before.
Yup. For instance, nyan-mode for Emacs has had a DOI for years now: http://doi.org/10.5281/zenodo.164185.
(No reason for doing it, I was just curious if I could.)
(No reason for doing it, I was just curious if I could.)
That’s a pretty interesting idea. Can you give me some thoughts as to where it’s such a paper should be published?
The Journal of Open Source Software https://joss.theoj.org is a nice example of this.
Perfect, thank you very much.
Sci-fi legend Kim Stanley Robinson was a keynote speaker at our virtual conference!! I guarantee that wouldn't of happened if we were in person.
Virtual conferences are terrific. I just helped host the AirMiners conference [1] on carbon removal on May 13. Because it was a virtual conf we were able to achieve top speakers, international attendees, and just seemed so much more effective. (KSR's keynote is here: https://youtu.be/9Pw0n0CeK0k)
It's going to be an uphill battle to get me to attend an in person conference again.
[1] http://airminers.org
Virtual conferences are terrific. I just helped host the AirMiners conference [1] on carbon removal on May 13. Because it was a virtual conf we were able to achieve top speakers, international attendees, and just seemed so much more effective. (KSR's keynote is here: https://youtu.be/9Pw0n0CeK0k)
It's going to be an uphill battle to get me to attend an in person conference again.
[1] http://airminers.org
This looks great! I’d love to see some of the fantastic context in your lovely Twitter thread included on the website.
As someone who’s currently in industry rather than academia, I interpreted “virtual conference” as “a Twitch or YouTube stream, and likely some sort of chat space like a Discord”. That landing page does a great job of explaining the technical underpinnings of your tool, but doesn’t actually explain what sort of site/services the web server actually serves!
Even just including the GIF from the beginning of your Twitter thread would be super helpful.
As someone who’s currently in industry rather than academia, I interpreted “virtual conference” as “a Twitch or YouTube stream, and likely some sort of chat space like a Discord”. That landing page does a great job of explaining the technical underpinnings of your tool, but doesn’t actually explain what sort of site/services the web server actually serves!
Even just including the GIF from the beginning of your Twitter thread would be super helpful.
Thanks for the comment! Updated the readme. During the conference we ran we did integrate chat and video tools (Rocket.chat, slido, slideslive). This is really just the glue to pull those parts together.
The conference was amazing, and the portal was very usable! I really liked the page where you can find similar papers from a clustering map. Thank you and the team for the hard work, top stuff!
The T-Sne view for papers is a killer feature. Loved it for ICLR, glad more will be able to use it. Thank You!
This is lovely. I like both the approach and the output.
I read the website and the readme, I don't understand what that is.
I'm a professor, not a software engineer, so happy for any comments or contributors!