Text Unit Classifer Suggestion Appears Empty

[dwmcqueen]

This may be a user issue, but I have classified several text units and the suggestion list continues to be empty. Is it functional?

[reddalexx]

Hi.

Text Unit Classifier Suggestions mainly depend on 4 things:

• how you marked text units using "class name" and "class value"
• which item you choose in "classify by" list (terms, geo entities, parties)
• classifier algorithm
• and confidence value

So doing a classification you say: let's classify all text units
using specific terms from text (geo entities, party names) like anchors and
specified class [class name] values [class value] as labels for class groups.
And then set classifier method and Max confidence value.
Consider to coordinate marked text units with selected "Cluster By" form field
and may be decrease "Max Confidence" value.

[dwmcqueen]

That makes sense - so it is functional. As I understand it, the "class name" for a similar text unit would be the same but the "class value" could differ, correct? For example, I could have a class name for "Assignability" but then several different values based on how the text unit reads? Then, it is important to have enough samples so the classifier runs?

[reddalexx]

Yes, you are right.
F.e. if a user wants to classify by geo entities,
class_name could be "Case1", class_values could be "northern" text units with geo entities like Alaska and "southern" for text units with Florida or Texas
and then run classification with classify_by="geo entities"

And yes, there should be as much as possible samples to make better classification.