Token removal/insertion in already annotated file strategy

[Agustin Dei]

Hi,

Which method do you suggest for removing, inserting, or modifying a token in an already uploaded and annotated file in Inception?

For instance, after having annotated a text we notice tht a token must modified, removed or inserted for different reasons (typo, variant of the text, etc), so we download the file in conllu and modify it locally. However, the file because unreadable in inception because of a change in ID indexing or relations.

Thanks

[Richard Eckart de Castilho]

Hi,

> On 19. Oct 2025, at 14:26, 'Agustin Dei' via inception-users <incepti...@googlegroups.com> wrote:
>
> For instance, after having annotated a text we notice tht a token must modified, removed or inserted for different reasons (typo, variant of the text, etc), so we download the file in conllu and modify it locally. However, the file because unreadable in inception because of a change in ID indexing or relations.

Editing tokens and sentences is not fully supported in INCEpTION yet.

You can enable the experimental support by adding these lines to your settings.properties and restarting INCEpTION:

```
ui.token-layer-editable=true
ui.sentence-layer-editable=true
```

See also: https://inception-project.github.io/releases/38.3/docs/admin-guide.html#sect_settings_segmentation

When these settings are present, you can add Token and Sentence layers to your project via the (Add) button in the layer configuration. After adding them, you also have to enable them. On the annotation page, you need to choose an editor which is not sentence-oriented, e.g. "brat (line-oriented").

If you delete a token or sentence, it should expand the previous or following token/sentence to cover the deleted area. To split a token/sentence, you would shift-click at the split location.

Mind that there are known and probably unknown problems that can appear when you start editing tokens and sentences. For more information see here:

https://github.com/inception-project/inception/issues/1778

So feel free to try it.
Be prepared to run into problems.
Let me know what problems you run into.

Cheers,

-- Richard

[Agustin Dei]

Hi Richard,

Thank you for the quick reply !

I've added these lines today :

```
ui.token-layer-editable=true
ui.sentence-layer-editable=true
```

I'm doing a test with a copy of an already annotated text. So I deleted a token but as you announced it expanded the previous or following token. But splitting the new a token with shift-click at the split location would get me back to starting point. 

So 

1. I deleted the token from the Token layer keeping it added to the previous token, 

2. I downloaded the conllu file

3. I modified the conllu directly from vs code erasing the extra token (that had been added to previous one)

4. I re uploaded the file to inception and it seems to work, since it had all been re-indexed at step 1 !

I will do other tests with tokens being head of dependency relations to see what happens and let you know.

However, I am I afraid I still cannot modify the token: for instance a typo error directly from inception. I can only delete it. We should do this directly from a text editor?

Best,

A.

[Richard Eckart de Castilho]

Hi,

> On 19. Oct 2025, at 18:06, 'Agustin Dei' via inception-users <incepti...@googlegroups.com> wrote:
>
> However, I am I afraid I still cannot modify the token: for instance a typo error directly from inception. I can only delete it. We should do this directly from a text editor?

INCEpTION does not allow to change the document text.

If you are working with a tool that supports teams of annotators, there needs to be some common ground between the annotators.
That is what the document text is in INCEpTION.
If annotators were able to change the text at will, their documents would diverge and become incomparable with each other.

If you want your annotators to help you fix typos in the document text, then you could create an annotation layer which annotators could use to annotate typos with their corrections (there is an "Orthography" layer which might be used for that). You could then curate such corrections just like other annotations. Finally, you would need to write a script to apply the correction annotations to your actual text external to INCEpTION.

Cheers,

-- Richard