Haystack: Deep Annotation Tool Offsets on Windows

Created on 19 Sep 2020  路  2Comments  路  Source: deepset-ai/haystack

Describe the bug
When using the Labeling Tool "Deep Annotation" on Windows there were offsets within the labeled answers. When I selected an answer to a question and then viewed the answer in the "Export Labels" tab, the answer there differed from the text that I had selected.
For example: I selected "vor ca. 40 Mio. Jahren" and the answer in the Export Labels tab is "igoz盲n (vor ca. 40 Mio".
This behavior only occurred with the text files I created myself. When using other text files, e.g. files from the Haystack Tutorial 1 (Game of Thrones Texts), the labels were correct.

Solution:
After comparing the text files I created with the GOT text files, I noticed that the line break type from the GOT files were "LF" while the line separator in my files was "CRLF". So I simply changed the line break type in my text editor to "LF" and now everything works just fine.

System:

  • OS: Windows
bug

All 2 comments

Thanks for adding the detailed bug description @sophgit

We already created a fix that should convert windows, MacOS and Unix txt files into a single format. We are currently testing the changes on our staging server and will let you know once they are on the production system.

annotate.deepset.ai is updated and should be fixed.
Closing this now, feel free to reopen if the issue still persists.

Was this page helpful?
0 / 5 - 0 ratings

Related issues

rshtirmer picture rshtirmer  路  3Comments

zshnhaque picture zshnhaque  路  6Comments

Ierezell picture Ierezell  路  4Comments

tholor picture tholor  路  3Comments

antoniolanza1996 picture antoniolanza1996  路  4Comments