Input and output data

Input data

Tara supports the following types of input for Transaction Screening:

• Input data
  • Streaming API input
  • File input
  • Request Data Store
• Output data

Streaming API input

The system obtains input data from the built-in streaming service. You can define and update API endpoints. A Business Process (BP) is available even when all records are processed.

To use the built-in streaming service, do as follows:

1. In your BP, select the Data tab.

2. Check the Streaming Records from External Sources box and specify the external source in the field.

   

   For more details, refer to API | Tara.

File input

The file input feature allows uploading an input CSV file. A BP finishes when all records are processed.

• The file-in-original format contains a JSON request. This format contains the following columns for execution:
  • request_json_object (required)
  • decision_gold_json_object (optional)
• The file-in-flat format contains an unfolded request_json_object. A BP treats each row as a separate request.

To upload data, follow the steps:

1. In your BP, select the Data tab, and click Upload Data.

   

2. Click Add, find your CSV file, and click Open.

   

3. Wait till the file is processed and click Upload.

   

note

Using a CSV flat file as the template is the easiest input method for uploading input data for Transaction Screening.

Request Data Store

Tara accepts standardized input in JSON-formatted strings and saves requests into the pss_request_processing_v1 Data Store. For more details, see Run screening | Analyze results.

Example of the pss_request_processing_v1 Data Store:

Column	Description
system_id	Database generated ID
request_id	UUID generated before adding a record to Data Store
request_json_object	Original customer data for adjudication in the JSON format representing a Message object
decision_json_object	Model Decision in the JSON format representing a MessageDecision object
decision_gold_json_object	Gold Model Decision in the JSON format representing a MessageDecision object; used for model training and calculating statistics
message_json_object	Processed or enhanced copy of customer data for adjudication in the JSON format representing a Message object; data from this column is sent to the model for processing
request_insert_time	Insert data timestamp
last_updated_time	Timestamp of last update
decision	Final Message decision
status	Processing status; updated during record processing
client_id	User who sent the request for processing

Output data

The Tara BP's output data field is model_result.

The model sends a response with structured data in the JSON format:

• Category: reason code (differs for each client). The sample categories are as follows:

  "type mismatch"
  "type match"
  "name mismatch"
  "name match"
  "partial_name_mismatch"
  "location mismatch"
  "location match"
  "date of birth mismatch"
  "date of birth match"
  "msg names mismatch"
  "ss_name_screened_name_mismatch"
  "multiple names found"
  "blacklisted"
  "not implemented"
  "escalate"

• Explanation: human-readable decision explanation

• Score: classification model confidence of the alert to belong to the predicted topic

• Model-decision attribute: full model decision in the JSON format

A sample model's result is as follows:

{
  "score": 1
  "category": "RESOLVE"
  "estimatedAccuracy": 0
  "attrs": {
    "model-decision": {
      "messageId": "58198"
      "suggestedAction": "RESOLVE"
      "hitDecisions": [
        {
          "hitId": "8194965"
          "suggestedAction": "RESOLVE"
          "comment": "Name mismatch. Screened entity name 'Witting-Armstrong Inc.' does not match Watch List Entity name 'Terry, Konopelski and Auer Solutions Inc'"
          "reasons": [
            {
              "code": "type match"
              "description": "Screened entity type <organization> matches watch ..."
            }
            {
              "code": "name mismatch"
              "description": "Name mismatch. Screened entity name 'Witting-Arms..."
            }
            {
              "code": "location match"
              "description": "Screened entity locations <VUT> either matches or ..."
            }
          ]
        }
      ]
    }
  }
}