search

OHNLP / MedTator

A Serverless Text Annotation Tool for Corpus Development
https://ohnlp.github.io/MedTator/
Apache License 2.0
54 stars 20 forks source link

Unable to load schema #10

Closed shashank140195 closed 1 year ago

shashank140195 commented 2 years ago

When you save the schema which by default gets saved by the .yml extension, you cannot load it back. It is all empty.

hehuan2112 commented 2 years ago

Thank you for your feedback on this issue. By default, the schema editor should save the schema as .yaml format, so if the extension is changed to .yml, the current MedTator couldn't recognize it correctly. You can try to rename your schema file extension to .yaml for now.

And we just found that there is a bug related to the open schema dialog of the .yaml format. We have added this issue to our next release plan. We will fix the dialog, add the .yml in the supported extension list, and show some information for similar issues instead of just showing empty list.

shashank140195 commented 2 years ago

Thank you for the suggestion. Also, is there any way we can save the current state of the annotation? I have more than 1000 documents that need to be annotated. I accidentally closed the browser and all annotations are gone. So I am looking for a way to auto-save the annotations or the current state that can be loaded again.

hehuan2112 commented 2 years ago

You are welcome!

We agree that auto-save would be a very helpful function and we also got similar comments from other users. The current issue related to the auto-save is that the security strategy of web browser doesn't allow JavaScript to read/write local disk files without the user's manual intervention to each individual file. For example, when we open web browser and start MedTator, the MedTator JS cannot read user's local disk directly to load last state. The user needs to drag/drop/select files to MedTator to start annotation. And when saving annotation files, the web browser will show a dialog about write file permission for each file, otherwise the annotations cannot be saved.

As a result, it's hard to save annotation states automatically like Google Docs (it has backend server to save edits) or Microsoft Word (it can r/w local files directly). We are trying to make a semi-auto-save feature which may be helpful in reducing the clicks.

hehuan2112 commented 2 years ago

@shashank140195. For large corpus annotation (e.g., >1000 files), it is challenging to manage such large number of files. So in our practice, we can use the following two ways to track annnotation:

  1. sort files by number of tags. the number of annotated tags is shown in the file list, and the user can sort list by the number of annotated tags. So, by sorting the files, the unannotated files (i.e., number of tags is zero) can be display at top.
  2. label function. The user can add colored label to each file. For example, blue label for annotated files, yellow label for concerned files, green label for confirmed files, etc.

Then, each time when the user starts annotation, just drag the whole annotation file folder to MedTator and then can load labels to continue annotation. We are still thinking about how to improve this process for a better annotation experience. Please let us know if you have any thoughts.

hehuan2112 commented 2 years ago

Hi @shashank140195, we have fixed the schema open bug and added .yml in the latest release, which can be found at:

  1. Online version: https://ohnlp.github.io/MedTator/
  2. Standalone version: https://github.com/OHNLP/MedTator/releases/tag/v1.3.8

In addition, we added a new experimental auto-save function in the latest release. Please feel free to try and share your comments with us. :)