TLDR - Sysrev AI models are down. Fix in 1-2 days.
SR is growing quickly and user activity has outstripped our modelling service's ability to keep up. SR models train on over 5 million data points derived from sysrev screening and data extraction activities. These models update daily, and with the recent influx of new users (SR just broke 10k registered users 🎉 ) our hardware is just not able to keep up.
We expect this to be fixed in the next 1-2 days and we have some big announcements coming that will more permanently address this problem. Subscribe to this Linkedin page to learn more!
Biobricks will also provide document streams for the upcoming SRVC release.
Learn about open source modular review at https://t.co/ZQeAIUJSSR and open source data pipelines at https://t.co/hBzAtq5XIs.
I hate writing #bioinformatics code for extracting data.
bio is particularly bad, so much data, download links, apis and other stuff I don't care about. Why does every db need a web app?
Just give me the data.
That's what we're solving at https://t.co/ZCmuxf8RcM
The model now trains on 1125867 reviewed documents and 4527844 labels.
The models work on RIS file projects and any projects that involve reviewing text.
If your models aren't working for a project, please hit the "issue" button in the upper right of your project.
SRVC is alive! Check out https://t.co/ZQeAIUKqIp.
SRVC is an open source, modular, document review system designed to run anywhere and accelerate the integration and development of novel document review methods.
@drzhelnov Thank you, we will fix that. In the meantime you can use:
Sysrev: A FAIR Platform for Data Curation...
https://t.co/BmFGIUFGVf
And you might be interested in "Applications for open access normalized synthesis..." which also used sysrev:
https://t.co/o3GZr3XmpR
The reviewers included the article "Assessing Household Willingness to Pay for the Conservation of the Phou Chom Voy Protected Area in Lao PDR".
The models agreed with the reviewers on every provided label https://t.co/RWq8U6eiLJ
Cultural ecosystem services, monetary evaluation, forests/woodland and evaluation methods. A public sysrev for forestry lovers at https://t.co/J2LdzWAGb8.
The sysrev models work pretty well on this one. Here's the prediction histogram for the screening models.
models are now functional. The updated system:
1. works on pdfs (still testing accuracy)
2. uses fine tuning on https://t.co/pfbQqQs25a models
3. creates a 'label embedding' for every sysrev label
4. updates daily across all sysrev (~1.1m articles)
@Augusti67333911 Hi Augustine, we have emailed you the invite link. If you did not receive it, can you open your pms here and I will provide it? Thank you.
@Augusti67333911 Hi Augustine, thank you for the ping. It looks like that project is having an issue with the invite link. We will send it to you shortly.
https://t.co/07u6NveCsA has been receiving more labeling activity lately and our modeling system is having a hard time keeping up.
If the models for your project have not updated for a while, please be patient. We are in the process of updating our system.
If mechanism-based risk assessment is your bag, you should sign up for the next EBTC webinar! It should be pretty interesting, looking at key characteristics of carcinogens and data management issues in assessments. Pretty left-field stuff, trust me. 👍 https://t.co/uIkfwtqMDC
@amylibrarian We believe the transparency and reproducibility benefits of public repositories are quite significant.
Since #Sysrev stores all file imports and public projects are #openaccess, this is something we do by default (the exception are PDFs which have special rules).
While the Queen was having a party, I was diligently beavering away at a new general-purpose systematic review protocol template. If you follow it, it should ensure you are compliant with COSTER, PRISMA-P, ROSES, PRISMA-S, and provide PROSPERO data. https://t.co/4IlBiExAgs