Cannot edit cells of types "array of Entity"

[gbella]

In step 3 the "Edit string" button does not appear.

[DavidLeoni]

This is a 'feature'. An Entity [ ] might have long entity names. In this case, text in the column will be considered long text, and ODR will mark the column as 'original column'. For such columns there is no 'edit cell' button available. ODR will then ask you in step 4 to extract entities from the original column to create a new 'target column' with the entities.

Note if you really want, you can still do a transform on a whole 'original' column. Admittedly, all of this is confusing, and we should decide once and for all a policy about editing, see also #114

[gbella]

What do you suggest then? The current situation just doesn't let me fix the
entities.
Do we need to discuss?

On 29 February 2016 at 17:13, David Leoni [email protected] wrote:

This is a 'feature'. An Entity [ ] might have long entity names. In this
case, text in the column will be considered long text, and ODR will mark
the column as 'original column'. For such columns there is no 'edit cell'
button available, but if you really want you can do a transform on the
whole column. Going back to the Entity [ ] scenario, ODR will ask you in
step 4 to extract entities from the original column to create a new 'target
column' with the entities.

Admittedly, all of this is confusing, and we should decide once and for
all a policy about editing, see also #114
#114

—
Reply to this email directly or view it on GitHub
#185 (comment)
.

[DavidLeoni]

I don't understand what you need to fix - relational entity names should only help nlp finding the id for entities - but if you want to fix the name string manually you defeat the purpose of nlp...
Fixed relational entity names might be needed if we import them as new entities, but we're not going to do it any time soon, and if you want to fix the entire column with a transformation to remove say a common problem, you can still do it.

[gbella]

NLP is not the goal, it is only a tool. Sometimes it fails. When it fails,
the human needs to intervene to fix the mapping. Currently there is no way
to add entities manually into the newly created column. This is a serious
shortcoming that should be fixed shortly if possible. I understand that
this feature would be complex to implement and requires a new UI feature:
we need to insert a new entity reference into an array item in the newly
created column. Tell me if you see any sensibly straightforward solution to
this. This is really a feature that we need.

My idea was (I think we discussed it) to create in each empty array value
some kind of placeholder entity, e.g., with ID=0 and name "not found" or
something like it. This would make it a clickable link that can be edited
with the disambiguator popup. You just need to do a little bit of
post-processing at the end of step 4 to replace the placeholders where ID=0
by NULLs (or whatever you are currently using).

What do you think?

On 29 February 2016 at 17:36, David Leoni [email protected] wrote:

I don't understand what you need to fix - relational entity names should
only help nlp finding the id for entities - but if you want to fix the name
string manually you defeat the purpose of nlp...
Fixed relational entity names might be needed if we import them as new
entities, but we're not going to do it any time soon, and if you want to
fix the entire column with a transformation to remove say a common problem,
you can still do it.

—
Reply to this email directly or view it on GitHub
#185 (comment)
.

[DavidLeoni]

Solving the issue of this bug is more subtle as requires fiddling with ODR automations, which is quite a pain.

So, we can do as you say, just in another bug: #194

[DavidLeoni]

Notice that now you can do some array editing even in step 3 by using GREL with new commands described in #194