doccano / doccano/doccano

Annotation lags or does not show up in long documents

Open
#1,880 8 comments 8 reactions 0 assignees View on GitHub
enhancement
Dominant language
Python
Stars
10.8k
Forks
1.8k
PR merge metrics
No merged PRs in 30d

Description

We have a sequence labeling project where we annotate long documents for Named Entites(4-500 words). The documents are also relatively rich in entity mentions.

We have tested various versions of Doccano (1.7.0, 1.6.2, 1.5.5), from the official docker image for each version, on a remote server with Google Chrome 96 and Firefox 91 as well.

The issues came up with all tested Doccano versions, and they were similar across browsers, but not the same:
1. In Chrome, after about 60-70 annotated entities, if the annotator tried scrolling through the document to search for missed entity mentions, then selected the text to label it, the label did not appear. After scrolling past that particular annotation, and scrolling back, the annotation showed up. This is not just jarring for annotators, but results in them trying over and over, and having to delete a lot of excess annotations, which makes work much slower and more tedious.

![Here is how this looks like](https://user-images.githubusercontent.com/17028394/175944381-313492ec-a151-4885-a0ee-26601fa690b8.gif)

2. In Firefox the annotation does not cease completely, however the rendering of the text gradually becomes slower upon scrolling as the number of annotations increases, and eventually it won't even display any text whatsoever.

Upon a bit of a lookaround in the browsers' inspector we are suspecting that the problem comes from rendering the lines of text as SVG images.

In the example we used the Wikipedia article for the [Normandy landings](https://en.wikipedia.org/wiki/Normandy_landings), which might have changed since, so we are linking the text of the excerpt.
[normandy_landings.txt](https://github.com/doccano/doccano/files/8992032/normandy_landings.txt)

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.