Change records status based on user feedback

Hi

@merve

!

This includes user feedback in the record:

- Status: Flagged as correct can be used for retraining directly, the rest need to be reviewed (status=Default)
- Metadata: We store the feedback as metadata so we can filter by "Ambiguous" for example and focus on validating those records.

app.py

@@ -19,7 +19,12 @@ examples = [
   ["Lütfen yardım Akevler mahallesi Rüzgar sokak Tuncay apartmanı zemin kat Antakya akrabalarım göçük altında #hatay #Afad"]
 ]
 
-def create_record(input_text):
+def create_record(input_text, feedback):
+    # define the record status based on feedback
+    # default means it needs to be reviewed --> "Incorrect" or "Ambiguous"
+    # validated means it's correct and has been checked --> "Correct"
+    status = "Validated" if feedback == "Correct" else "Default"
+    
     # Making the prediction
     predictions = nlp(input_text, aggregation_strategy="first")
     
@@ -40,6 +45,8 @@ def create_record(input_text):
         tokens=words,
         prediction=prediction,
         prediction_agent="deprem-ml/deprem-ner",
+        status=status,
+        metadata={"feedback": feedback}
     )
     print(record)
     return record
@@ -59,7 +66,7 @@ class ArgillaLogger(FlaggingCallback):
     ) -> int:
         text = flag_data[0]
         inference = flag_data[1]
-        rg.log(name=self.dataset_name, records=create_record(text))
+        rg.log(name=self.dataset_name, records=create_record(text), flag_option)