Using document data to enhance models

If you have Classification Server installed and configured within Studio, it is possible to use data from your own documents to enhance a model help within KMM. To do this, you first need to upload those documents using the Text Analytics component.

Uploading documents

Uploading the document set

The document sets used for the text analytics are managed from Semaphore Studio’s homepage. From Document Analyzer (DA), click the three little dots:

From this link you will be able to create and manage document sets.

You create new sets by dragging and dropping the files you want to use into the left hand pane:

Once this is done, the document set is available:

Using uploaded document sets

Within KMM navigate to the term you wish to enhance and select the Text Analytics side panel.

You are asked to select the labels from the current concept in the analysis, the Classification Server instance, then the Document Set.

Once you have selected these, click the “Analyze” button.

You will then be shown the results of the analysis.

On the “Results” pane, you are shown for each document any instance of the selected label, and signficant noun phrases that either overlap it (these might be useful for more precise matching, or for preclusion), or are near to it (these might be useful for better evidence matching).

On the “Proximate Candidates” panel you will see an ordered list of promixate phrase found across all documents - ordered by the number of times they appear across all documents and in how many documents. Overlapping candidates are similarly shown on that pane.

Clicking on “Add Candidate” on the lists, or clicking on a highlighted phrase will add the selection to the right-hand side of the text analytics panel.

Clicking “Add as a label” will allow the user to select the label type and the language of each label before adding them in as labels of the current concept. Note, you can edit the label value before pressing submit if the values need tweaking.

Clicking “Add as concept” will allow the production of multiple narrower concepts. Note that the parent of the new concepts should be specified. By default this is the current concept, however, you can use the Search as You Type interface to another concept or even a concept scheme.

Viewing the results in Document Analyzer

By clicking on on any document link in the “Results” view, you will be taken to that document in the Document Analyzer. Here the entire document will be visible and the candidates viewable in full context.

Here too, you can select “Add +” to add the candidate to a set to be added. Clicking on the “Add As” button at the top will then allow you to select whether to add as labels or concepts