[IX] Support metadata extraction from text fields as source

[aphilop]

We need to expand the currently implemented functionality of metadata extraction with the following requirements:

• A metadata extractor can have as source either PDF documents or text fields.
• Only one source can be selected when creating a metadata extractor.
• The source of each extractor (PDF or selected text property) is indicated in the relevant column (SOURCE) in the extractors view.
• When reviewing a metadata extractor that has as source a text field:
  • The functionality (table view, information display, interactions, stats & filters) remains the same as currently implemented.
  • The review panel should display on the top part the relevant text field while the lower part remains the same as we have it now for the different types of properties.
  • The click-to-fill button should perform a "copy paste" function where you select some text, select click-to-fill and this puts the value in the relevant input filed. Under the hood it has no effect larger than just typing in the value in the input field.

[aphilop]

@RafaPolit
As discussed, these are the high-level requirements for extending the metadata extraction to support text fields as source.
Also listed are the relevant issues where the UI changes are illustrated.

Please review and let's discuss them in the next product call to address any questions/suggestions and make sure the backend team has everything is needed to start implementation. Thanks!

[aphilop]

@huridocs/backend to talk with ML and frontend team to discuss the contracts.

[aphilop]

Putting this for the next sprint with the following expected outcome:

• Clarify the required contracts: ML, backend, frontend
• Create relevant issues for the backend development

[aphilop]

When creating a new extractor, after waiting some time the system returns an error:

However, when refreshing the new extractor appears in the list:

[aphilop]

Update: this is resolved as there are double entities in the library so this is not a problem in IX.

Entity entries are double-listed in the table:

[aphilop]

Triggered the "Find suggestions" process for a select property with 19596 entries in the extractor.
The process somehow stuck in the "Training the model" step. After quite some time, I got out from the extractor and got in again and then 19596 entries were marked as obsolete and 3 as mismatch.

[aphilop]

Triggered the "Find suggestions" process for a select property with less entries (1212) in the extractor.
It delivers back mostly obsolete suggestions. Rerunning the process does not seem to fix the problem.